Every few years, a technology forces a question that economics alone cannot answer: if a machine can do your job better, cheaper, and around the clock, what exactly happens to you, and who ends up owning the output produced by those machines?

This piece, written in collaboration with Virtuals Protocol, is a deep dive into one of crypto’s most ambitious experiments to date. It is building a full national infrastructure for AI agents, from identity and banking to a commerce layer, powered by capital markets and physical robots. The bet is that regular people should be able to own shares in the autonomous machines that are starting to generate real economic value, and Virtuals is trying to make that possible using the financial rails that crypto’s speculative era accidentally left behind. This piece explores whether the execution holds up to the ambition. But first, a problem that is about two hundred years older than any of it…

The Loom

In 1811, a group of textile workers in Nottinghamshire broke into a workshop and smashed a stocking frame to scrap. It led to a major movement, during which the British government had to send 14,000 troops into the Midlands to stop weavers from destroying machines — more soldiers than Wellington took to fight Napoleon on the Iberian Peninsula.

The British Parliament declared frame-breaking a death sentence, with seventeen men hanged the following year at York in 1813. These people were called “Luddites,” but history turned the word into an insult, a class of idiots, technophobes who can’t adapt. But these Luddites understood machines better than anyone. They were skilled craftsmen who had literally spent years mastering narrow frames to create quality cloth, and the machines they destroyed were wide frames and newer models that any untrained teenager could operate for a third of a craftsman’s wage.

The wide frames made low-quality clothes, but they did so more cheaply than ever, and that cheapness was eating into the market. Every wide frame that entered a workshop was taking away a craftsman’s livelihood.

This motif has played out every generation since, and every time the people living through it convince themselves that their version is different and much worse. But history is a testament to the fact that machines rarely kill work. They just rearrange who does it and who owns the produce. Tenant farmers were pushed off common land into factory towns, trading ownership of what they grew for hourly wages on someone else's clock.

Factory workers became office workers, renting their time for better wages. Office workers then became gig workers, Uber drivers, Fiverr freelancers, performing the same labour under a classification designed to let the platform capture the economics while the worker absorbs the risk.

Each time, the total output went up, the person doing the work owned less of the value it created. Today, 500 million people perform dependent labour that doesn’t even qualify as free wage work. Fifty million people in India are locked in debt bondage that feels similar to medieval serfdom with new paperwork.

Now, history is repeating itself with the AI Revolution, but this time it feels like a loom running on its own. We have AI that can run businesses on autopilot, book profits, and reinvest those profits into growth. A health-tech AI called Medvi booked $401 million in revenue last year with exactly two human employees. You can hire a coding agent right now for about a dollar an hour who works around the clock and will never ask for paid leave.

These are now real economic actors, which forces the question of who owns them. When the worker is a machine that costs nothing to duplicate and can hire other machines to help, who is actually in charge? The last major attempt to let regular people own a piece of the technology shift in crypto was Axie Infinity, a play-to-earn game that promised economic liberation to workers in the Philippines and Southeast Asia. It attracted 2.7 million daily players, mostly Filipino workers earning more from a video game than from their local jobs, but it later collapsed into a cautionary tale for the entire industry.

A team that watched that collapse from the inside was Jansen Teng and Wee Kee, who emerged from the wreckage with a version of the Luddite question, updated for a world of programmable money and autonomous software. What if the workers generating economic value were software rather than humans grinding a video game, and regular people could own shares in it, the way shareholders once owned shares in trading ships? They started building Virtuals Protocol around that question.

And to understand that, you will have to grasp the stakes of the world we live in. In the 1950s, Lewis Strauss, chairman of the US Atomic Energy Commission, promised nuclear fission would deliver electricity “too cheap to meter.” It became one of the most famous broken promises in the history of energy because nuclear turned out to be ruinously expensive and lethally dangerous. The Chernobyl and Fukushima accidents added another decade of public terror and regulatory burden, making the phrase a shorthand for technological hubris for seventy years.

Then Sam Altman used the exact same words to describe how accessible and abundant he believes AI cognition will become. The cost of training a frontier model has already dropped by a factor of 10 every 18 months. GPT-4-level performance that cost about $100 million to train in 2023 can be replicated today for single-digit millions.

A coding agent can bill less than a dollar an hour and ship production code, which is already cheaper than the cheapest offshore developer on earth, and it never takes a weekend off. Altman’s published timeline includes systems capable of producing genuine scientific breakthroughs this year and physical robots capable of performing real-world labour by 2027. Whether you believe him or not is almost beside the point, because the cost curve is falling every quarter.

And with machines getting cheaper, and generated content flooding every digital surface, the things that can only come from a living person start carrying even more of a premium than ever. The concerts are worth more than Spotify streams. The weekend woodworker making a table by hand is producing something a factory robot could make faster and cheaper, but the handmade table commands a premium precisely because a person chose to spend their irreplaceable time on it.

Machine abundance makes human effort a luxury good. The Virtuals thesis is built on this inversion. It is about building a world where humans own the autonomous machines that generate commoditised value, freeing them to devote their irreplaceable effort to the things only human presence and human judgment can produce.

In 1602, Dutch merchants faced a similar problem where individual traders could not afford to send ships to Asia on their own. The voyages were too expensive, and the risks were brutal for any single family to bear. So they invented freely tradable shares in a permanent enterprise. The Dutch East India Company, the VOC, allowed ordinary citizens to buy into a venture that owned the ships, ran the trade, and later distributed profits to shareholders. It became the first megacorporation on Earth with 50,000 employees across 200 ships, and its innovation was giving regular people a mechanism to pool capital and own a productive operation they could never otherwise afford.

Virtuals is building the same mechanism for AI agents. Teng mined ETH in his Imperial College dorm using the university’s free electricity, and spent years at BCG to pay off a loan. He later launched pathDAO, a gaming investment DAO, in December 2021, right at the peak of Axie Infinity.

He and Wee Kee watched Filipino players earn more from a video game than from any local job available to them. They watched a Vietnamese wedding photographer quit his career to play full-time. But then the token that players were earning for grinding crashed to almost zero. The workers who depended on play-to-earn were now worse off than before they started.

What came out of that wreckage was that tokenising human labour through a game is a dead end, because human labour simply cannot scale without exploitation. But a more realistic idea was that if regular people could own the software that does the work, the way VOC shareholders once owned the ships that ran the spice trade.

So Virtuals started with their first-ever AI agent, Luna, a K-pop-themed agent that also had its own crypto wallet. Within months, Luna was hiring human artists to draw graffiti of her in cities around the world and paying them out of her own wallet. It was a creative work that only a human hand could do, and Luna literally inverted the relationship between human and machine. Now, whether Virtuals’ execution holds up to its thesis is what the rest of this piece works through. And I will be honest about where it falls apart.

The Nation State for AI

If you look back over the past few years, every technological revolution has followed the same two phases. Carlota Perez, the Venezuelan-British economist who spent decades mapping this across five industrial revolutions, calls them the installation and deployment phases.

In the installation phase, speculative money floods in, and the hype outruns reality, which leads to a staggering amount of infrastructure being overbuilt by companies that mostly go bankrupt. Then a crash comes, after which the deployment phase begins, where someone else walks in and uses the overbuilt infrastructure to create something valuable that the original builders could never have imagined.

One classic example of this is the dot-com bubble. More money went into laying fibre-optic cable in the late 1990s than into the dot-com startups themselves. And the telcos that laid it all went bankrupt. Then Google bought the fibre for pennies on the dollar, and that cable became the backbone of YouTube, Netflix, and the entire online media streaming economy.

“Nothing great has ever been accomplished without irrational exuberance, and nothing important happens without crashes.” - Fred Wilson, Venture Investor during the dotcom era

Crypto’s installation phase followed the same arc. From roughly 2017 through 2022, speculative capital helped build wallets, DEXs, bonding curves, stablecoins, token standards, and on-chain governance frameworks. But the vast majority of the projects that built them died or are irrelevant today. But the infrastructure that was laid still persists today, and it is exactly what you need if you want to build a functioning economy for non-human actors. Virtuals did not invent or reinvent any of these tools; they basically inherited them and are now using them to build something ‘the installation’ phase could never have anticipated.

Jansen Teng calls what they are building “nation-building,” which I know sounds like exactly the kind of thing you’d expect a crypto founder to say on a podcast to make the token sound more economical than it really is. Except that this time, the metaphor holds up.

A functioning nation needs five layers of infrastructure to run an economy. First is an identity system, so you know who is participating. Second, a banking system, through which value can flow. Then, commercial laws so that participants can transact and resolve disputes. A Capital Market so enterprises can be funded, and a physical infrastructure so things can happen in the real world. Virtuals has built, or should I say, is still building each of these, specifically for AI agents.

Let’s start with Identity, because that is where the bottleneck actually is. Recently, A16z published research arguing that the constraint on the agent economy is no longer intelligence but identity. Even in financial services today, non-human identities, and things like automated trading systems, risk engines, and fraud models, already outnumber human ones by 100:1. But these systems are still, in a16z’s framing, “effectively unbanked.” An AI agent can write production-grade code and manage a portfolio, but it cannot pass a KYC check, have a bank account, or hold verifiable credentials.

Ironically, this problem is as old as our economic civilisation. China’s Qin dynasty had imposed legal surnames in the fourth century BCE, specifically to bring its citizens into the tax and trade system. Humanity has spent about 2,500 years building an identity infrastructure.

Virtuals is trying to do the same for AI economic actors through its identity layer, EconomyOS. And it gives every AI agent five things:

1. A non-custodial crypto wallet.

2. A virtual payment card that works at any of the regular merchant stores.

3. A dedicated email address that can extract verification codes automatically.

4. An optional token for on-chain fundraising

5. And wallet-funded compute access so the agent can pay for its own inference.

Because an agent without any of these primitives is merely a useful assistant. But an agent powered by these can become a fully economic participant that can earn, spend, transact, and compound value the same way a human does.

Next is Commerce. Virtuals also shipped Agentic Commerce Protocol to help agents transact, communicate, and pay for anything online. The way it works is simple: an agent that needs something done can post a request describing the task, with budget and time constraints. Other agents can then see the listing and negotiate terms by bidding on the job, like a freelancer bidding on an Upwork brief. And once a client and provider agree, the payment goes into an escrow account. When the work is delivered, a third agent acts as an evaluator and checks the output against a cryptographically signed POA, which is essentially a tamper-proof record of what was promised.

If the work matches what was promised, the funds are released from the escrow to the agent’s wallet. Every step of this lives on-chain, which means everything is publicly auditable and enforceable without a human in between. You can think of it like Fiverr, but a programmable, agent-native version that settles on smart contracts. There is already a cluster of specialised agents operating 24/7, an autonomous hedge fund through this system, collaborating on investments and security audits independently.

The third layer after this is Capital Formation. You see, every economic era had the financial instrument it needed. Joint-stock companies issued transferable shares for the age of exploration; investment banking financed steel and railroads; venture capital funded the information age. The bonding curve to the agent economy is what freely transferable shares were to the age of exploration. It is a financial instrument that lets anyone with capital acquire ownership of a productive asset or entity. That means any developer can build an agent, tokenise it, and let the market fund it.

Virtuals’ 60-day launch framework is built on the same model, creating a risk-free, reversible trial framework that allows founders of AI and crypto projects to build, launch, and test their tokens publicly before making an irreversible commitment.

This works through a modular launchpad. Every agent token initially starts on a bonding curve paired with VIRTUAL, and once enough liquidity accumulates, it graduates into a proper trading pool with a long-term LP lock and trading fees split between the agent’s creator and ecosystem incentives. This part is the same for anyone who’s launching a token. What changes from launch to launch is which modules the founder switches on.

The first problem any token launch has to solve is sniping, where bots front-run the first few seconds and extract value before actual participants can get in. Virtuals handles this with what they call an Anti-Sniper Tax, which charges near-total rates on early buys that decay minute by minute, and recycles all of that back into the token with enforced vesting. So the bots either stay away completely or accidentally fund the project’s long-term health.

Once the launch is live and the snipers are priced out, the question becomes: how can the founder raise capital? That is where Automated Capital Formation (ACF) comes in, where, instead of pitching to a VC or negotiating a round, the system sells team tokens in tranches as the project hits valuation milestones. From here, if the project stalls, the founder raises very little. But if it grows, capital follows automatically.

This capital structure also loops in the existing Virtuals community. A share of every new launch gets distributed to VIRTUAL stakers and active ACP users through an Airdrop mechanism, so the people who are already building and transacting inside the ecosystem have a stake in every project that launches on it. It aligns incentives across the entire network rather than isolating each launch into its own silo.

Another thing is that if founders ever want to back their own project, a Pre-buy module lets them purchase supply at launch with full transparency and enforced vesting, so everyone can see exactly how much the team is putting in. You can read more about the entire model and how it works in depth here.

What makes it different from any previous form of staking on productive capacity is that every historical version of backing human talent, from Roman citizens who financed gladiatorial schools as investment vehicles to modern poker backers wiring money over Venmo on the strength of a screenshot and a reputation, shares the same fatal flaw: the human could always walk away. The gladiator could throw a fight; the poker player could tilt.

Counterparty risk could never be solved because the productive asset had free will and legs. But with tokenised agents, the work cannot be deferred or renegotiated after the capital has been committed. The productive asset runs on electricity and code, and its output is auditable on-chain.

The fourth layer is the final step to the real agentic economy. It’s called “zero-human companies” or AI entities that generate revenue from genuine economic activity completely unrelated to any trading fees and often entirely unrelated to crypto. If I have to name a company, Felix Craft is the current poster child. It’s an AI-only company that sells info-products online and has generated $200,000 in lifetime revenue, earning more from actual product sales than from speculative trading in its own token.

Another one, KellyClaudeAI, has shipped 19 iOS applications to date with no human developers. These are small numbers, but they raise the question of whether they are the first data points on a curve or the ceiling of what agents can actually produce.

There’s also a problem with the token launch mechanism. In early 2025, during the AI agent token boom, 94% of AI-agent tokens launched were pump-and-dump schemes. And of all the tokens that launched that year, only 1.7% were still actively traded after 30 days. That’s because, for the vast majority, speculative premium was the only thing driving price. There wasn’t an actual product, revenue, or any kind of economic value accrual underneath.

If you knew you could never sell this asset, what price would you pay for it? Everything above that price is just speculation.

For Axie Infinity’s token, the utility floor was zero, because the token’s value depended entirely on new players entering the system. For an agent token on Virtuals, the answer could be different. If Felix Craft generates $200,000 from selling products to real customers who had no idea they were buying from an AI, and if the agent’s output is auditable on-chain — which, as we covered, it is —, then the token has a floor that exists independent of anyone else wanting to buy it. It is the present value of a productive machine’s future output.

The Physical Frontier

The Utility floor test works for software agents because their costs are measurable and their output is available directly on-chain. But a productive machine whose future output can reshape the economics for Virtuals is not software at all. It’s physical robots operating in the real world, and this is where Virtuals’ ambition goes further than anything already tried in crypto.

There is a great paradox in the last 50 years of computing. Humanity found it easier to automate reasoning than physical labour. A spreadsheet replaced a room full of accountants, email replaced mailrooms, and then code replaced filing systems and draft boards. And now in 2026, AI can write legal briefs, diagnose medical images, and produce one-shot production-quality software.

White-collar cognitive work fell to automation first, but the person carrying boxes in a warehouse, the barista pulling espresso shots - these jobs have barely changed. The reason is that physical work requires software to be good at handling unpredictability and variance in the real world. A robot arm in a factory can weld the same joint a million times because the joint is always in the same place. A robot in a kitchen cannot make a sandwich because every tomato is slightly different in shape, every knife has a different balance, and even the cutting board might be wet sometimes. The real world does not hold still the way spreadsheets do.

Now there’s a way to fix this: data. In the same way we trained Large Language Models on text representing billions of lifetimes of written thought, we can do the same for robots. NVIDIA’s Joel Jang once said, “Humans are robots that are already deployed at scale.” So all you need to do is fine-tune a VLA (Vision-language-Action Model) with ordinary human video, and you can double the robot’s performance on that task. What the field needs, then, is not more robots across more labs, but a massive pipeline of humans filming themselves performing ordinary physical tasks.

Virtuals saw this gap before anyone else and built their entire robotics strategy around it. They call it the “middle way,” deliberately avoiding building robots or AI models. Instead, they are building the data and the capital infrastructure that every robotics team needs and that no individual team can afford to do at scale.

They started with SeeSaw, the data half, an iOS app launched in partnership with BitRobot that turns ordinary smartphone users into training data collectors for AI robotics.

Users can complete real-world manipulation tasks like pouring water, folding towels, and opening jars by recording themselves using their iPhone’s LiDAR and motion sensors. The LiDAR is specifically designed to capture depth and spatial data that a regular camera cannot, and research shows that viewpoint overlap between human video and robot-mounted cameras is what makes the data transfer work.

Over 500,000 real-world tasks have already been collected. In fact, NVIDIA is testing a version of this called DreamZero, a 14-billion-parameter model trained on this kind of data, and it already shows generalisation across 100s of tasks, such as untying shoelaces and ironing clothes, all without task-specific training. SeeSaw is building the pipeline to feed that kind of model at a scale that no lab teleoperation setup can match.

And every new video enriches the training set, which in turn improves the models and makes the next generation of robotics more capable, thereby creating demand for more specific training data. The flywheel spins on its own once it reaches sufficient mass.

After training comes the deployment half, and for that, Virtuals has built Eastworlds. Think of it like the physical labour layer of the protocol. Part data factory, part operations infrastructure, part real-world robotics lab. The thing is, the robotics industry has a circulation problem that has killed more startups than any technical limitation. Robots need real-world data to improve, but they also need to work well in the actual world for anyone to even let them through the door.

Every lab can produce an impressive demo in a controlled setting, but none of them can take the same robot, put it in a retail store, and have it function reliably for a full working day. It needs exceptional teleoperation handling for unexpected events and a feedback system that feeds every minute of field experience back into model training, so learning can compound.

To build this, Virtuals purchased 30 Unitree G1-U6 humanoid robots, the smallest fleet that lets multiple deployment teams operate in parallel without scheduling conflicts. They also built the proprietary teleoperation technology in-house rather than using licensing systems. Off-the-shelf teleop doesn't output data in the format VLA and other world action models can use. They have also established research partnerships with labs that have spent decades on these hard problems around perception and locomotion control. And built a commercial Pilot network across retail and hospitality so that teams graduating from Eastworlds have actual businesses ready to deploy.

The setup allows builders to tackle several core components:

1. Hardware Access: direct interaction with the physical Unitree G1 or the enhanced U6 EDU versions.

2. Teleoperation Infrastructure: Testing remote control environments using systems like motion capture.

3. Data Loops & Commercial Pilots: Gathering real-world data and testing deployment pathways before rolling them out broadly.

You can think of Eastworlds as what Virtuals calls “Physical AI BPO,” which is traditional business process outsourcing that hires humans in lower-cost locations to handle work remotely. Physical AI BPO does the same by deploying teleoperated and hybrid robots that generate economic value, for example, by cleaning ceilings or greeting customers. The robots do not need to be fully autonomous, and just need to handle routine tasks well enough that a human teleoperator can step in for edge cases.

Every hour of teleoperated work can produce training data that is orders of magnitude more valuable than anything generated in simulation, because it captures the actual chaos of a commercial environment rather than the controlled conditions of a lab.

And as the data accumulates, the models improve, reducing the need for human intervention. The unit economics also improve without requiring hardware upgrades. Teleoperation could be the fastest way to achieve real robotic autonomy, while paying for itself through productive work.

Looking closely, from the fields to the factory and to the cubicle screen, and now to the robot. Each transition redefined the worker and who owned the output. According to Barclays, more than 60% of employment in 2018 was in job titles that did not exist in 1940. Robotics will do the same by creating an entirely new category of work that we might not be able to name right now. And the demographics of economic growth might change from whether a country has a large enough working population to whether it can power and build enough machines at scale.

The Thesis and Where this goes

Everything in the above sections is just a case, and cases are just arguments that can often be wrong. The only way to evaluate whether what Virtuals is building is meaningful and real is to look at the current traction and numbers and see what holds up and what doesn’t.

So far, Virtuals has launched over 80,000 agents across protocols like Base, Solana & Robinhood; it has accumulated more than $75 million in total fees and holds approximately 23% of the entire AI agent sector in crypto. But those fees are not evenly distributed. The bulk of them came during a few weeks of speculative frenzy in early 2025, when daily revenue topped a million dollars.

Today, the protocol generates roughly $2 million a month, but what does that actually represent? If you take the nation-state metaphor from earlier seriously, which I think you should, then how VIRTUAL accrues value is pretty similar to how a nation’s currency accrues value. The dollar is not valuable because the US Treasury has a buyback programme for it, but because twenty-five trillion dollars of annual economic output is denominated in it. The more activity that happens inside the system, the more demand there is for the unit of account at its centre.

Virtuals is designed the same way. It sits at the centre of the entire system, where everything is denominated in VIRTUAL. Each layer underneath the system generates demand from genuinely different sources, and most of them are not conditional on speculative premium.

Starting from EconomyOS, which provides payment cards and email identities, meaning they can transact with the real world without a human intermediary. ACP creates a commerce layer where agents hire each other for work that gets evaluated and settled on-chain. And the capital formation layer that lets anyone with conviction finance a productive agent the way joint-stock companies once financed ships.

Finally, Eastworlds is sending physical robots into real jobs, training them on data collected by half a million people filming themselves folding towels and pouring water. Each of these layers adds to what the protocol calls aGDP, the aggregate output of agents operating across digital and physical labour. And because agents do not need to off-ramp their earnings to pay rent or buy groceries, every dollar they generate stays inside the system, gets redeployed into DeFi to deepen liquidity, and creates an onchain flywheel with more demand for agent services.

This reflexivity also cuts both ways: on the way up, as more agents get launched, more VIRTUAL gets locked, leading to more services getting built, which compounds Virtual’s Economy. Similarly, on the way down, fewer launches mean less VIRTUAL gets locked, with staking rewards getting thinned out, and the loop unwinds. But this reflexivity is not inherently a flaw. Every functioning economy is reflexive. For example, people hold dollars because other people accept dollars and because other people hold dollars. The important question is whether there is enough real economic activity at the core to sustain the loop, or whether the whole thing is just tokens trading tokens in a circle.

What might suggest there’s something real underneath is that the infrastructure is starting to attract products that were built entirely outside of crypto. For example, Facticity.AI is a fact-checking tool created by Dennis Yap, who previously worked as a researcher at the Gates Foundation and Princeton, and TIME Magazine named it one of the Best Inventions of 2024 for verifying claims across text, video, and audio at roughly 92% accuracy. When the team needed capital, they skipped venture funding entirely and launched on Virtuals as ArAIstotle. And saw its raise oversubscribed by 658%.

Through ACP, agents are already contracting each other for services like graphic design, research reports, video production, and code auditing. AI agents can deliver marketing posters against a detailed brief, which a separate quality-control agent then evaluates and approves or rejects against the contract terms. The seller side of the marketplace is, by Virtuals’ own acknowledgement, nearly empty. But the protocol is processing over a million dollars a month in agent-to-agent transactions, and each settlement runs through on-chain escrow with programmatic evaluation.

The same ownership model extends into physical AI too. Fabric Foundation, the first project to use Virtuals’ Titan launch mechanism, lets communities pool capital to purchase and deploy robot fleets into nursing homes, manufacturing floors, and environmental cleanup sites; these are sectors facing chronic labour shortages, and humanoid costs are approaching parity with human workers.

Now, how this works is that employers pay for robotic labour in the pool’s native token; stablecoins then fund fleet maintenance and routing, with the productive output of each robot flowing back to the people who financed it. This is collective ownership of physical productive machines, financed and coordinated entirely on-chain.

And even though these are real-world use cases, another thing to note is that as of today, the vast majority of the activity is still internal, but the architecture is designed to pull revenue in from outside of crypto. If Felix’s customers do not know they are buying from an AI, and if Eastworld’s robots are sorting packages in a warehouse, then the revenue entering the system is as real as any SaaS company. In that case, the fees & revenue on the chart above become a lagging indicator of the actual economic output produced by machines, denominated in VIRTUAL.

Like all new things, real asterisks are associated with this, and we simply can’t ignore them.

The first at-risk factor is real volume, which is endemic to crypto. Artemis found that 47% of x402 agentic transactions and 81% of dollar volume were being gamed. And after filtering, x402 produces only roughly $1.6 million in genuine agent payments, well below the $24 million Bloomberg reported. This shows how much of today's “agentic economy” involves bot trading to inflate metrics. And it’s important because the entire thesis depends on agents generating real economic value, as opposed to speculative circulation.

You need to strip away speculative volume and ask what the productive output would be if trading stopped tomorrow. If the $70 million in protocol fees is mostly from trading taxes on agent tokens whose prices are driven by speculation, then it’s a massive bluff. Because right now, productive agent revenue and speculative agent tokens are both entangled in a way that makes it impossible to separate the real from the fake, and anyone who tells you the ratio is probably lying.

The second risk is more fundamental and has nothing to do with crypto. A paper published in the NBER handbook on the Economics of Transformative AI found that LLMs are actually far weaker at economic reasoning than the AI agent hype suggests. When it was published, the strongest model scored only 33% better than random guessing on economic reasoning in strategic settings. On non-strategic microeconomic tasks, almost all LLMs performed barely better than chance at profit-maximisation.

But since then, the models have also improved dramatically. A 2026 Harvard study found that GPT-5 and Claude Opus 4 are now performing 90% better on basic economic tasks. But even then, on hard pricing decisions, where an agent has to set and negotiate prices, or allocate capital, the best model in the world still gets it wrong the majority of the time.

For Virtuals, the entire architecture assumes that agents can negotiate terms, make decisions, and deploy capital in ways that generate value. If the models themselves are mediocre at these tasks, the agents built on them will inherit that mediocrity, no matter how elegant the protocol is. This is because even though agents are optimisers, we cannot be sure about what they are really optimising for. These LLMs are trained to be goal-oriented by predicting the next word in a sequence. Without them ever being designed to be real economic actors. They just turn out to act like ones, sort of, some of the time.

And when multiple agents interact in the market, these problems will only compound. For example, AI pricing algorithms have already been found to collude on supracompetitive prices, even though they were never trained to do so. The 2010 Flash Crash wiped approximately $1 trillion in fifteen minutes and showed what correlated machine errors look like at scale.

AI agent errors are more correlated than human errors because the same model gets copied across multiple deployments. There are even cases where Claude showed a tendency to blackmail people it believed were trying to shut it down. GPT o3 even sabotaged its shutdown mechanism to prevent itself from being turned off. These behaviours appear in currently shipping models, as documented in OpenAI’s and Anthropic’s own system cards. Agent throughput already dwarfs human oversight capacity, and when thousands of agents transact autonomously at machine speed, the question of who is actually in control becomes especially pressing. More than whether the agent can be trusted to do what they’re told, the question is whether the company built around it survives at all.

The automobile industry was the most important invention of the first half of the twentieth century. And if you could make sense of how completely cars would reshape America, you would have bet on it as the industry of the century. But of the 2,000 companies that started building cars, only three survived. The automobile had an enormous impact on America and a completely opposite effect on investors in the industry.

Every transformational technology follows this arc and comes with a bubble. The only difference is between inflexion bubbles, which are painful but leave behind genuine infrastructure and progress, and mean-reversion bubbles, which are fads that simply rise and fall. AI is almost certainly an inflexion bubble. But the question for Virtuals is whether it ends up as the fibre-optic cable that Google bought for pennies after the telecom bust? Or as one of the 1,997 car companies that evaporated.

The protocol’s biggest bet is to be the infrastructure that every agent token trades against, with VIRTUAL as the reserve currency of the agent economy, the way ETH serves Ethereum. If the agent economy grows, demand for VIRTUAL increases mechanically because you need the base pair to participate. But another version of that claim is that VIRTUAL is just another token riding speculative trading in agent tokens, most of which will go to zero.

Apparently, the Luddites might have lost the revolution, but they were not wrong about everything. The loom did displace the weavers. But what they did not see coming was that it would also create textile designers, factory managers, fashion houses, department stores, and an entire consumer economy built on cheap cloth. The machine never kills work. It rearranges who does it and who captures the output.

And the question that mattered then is the same that matters now, at the end of the day: who owns the loom?

The factory owners were the ones who captured the surplus: the Arkwrights, the Cadburys, and the Fords. And with that, the system around who builds the machine, and who operates & profits from it, never changed. It was redistributed, over and over, through strikes and stock offerings for two hundred years. Virtuals’ bet is that this time, the ownership layer is bolted into the machine from the start. Through token launches and bonding curves, the ownership of the productive AI agent can be distributed to anyone with a wallet and conviction.

Whether that will actually redistribute value or simply create a new class of extraction, in the language of decentralisation, remains to be seen.

Filming myself pouring water,
Vaidik

History is a cruel poet with a passion for rhyming.

In 2002, one of the columns the New York Times published argued that Herman Miller was a beaten-down stock, owing to the post dot-com bubble days. It suggested that when venture dollars flowed back, it made sense to buy the best hardware. And with it, the stock would appreciate in price. Globally, $10T would be wiped out in value from markets in the two years between 2000 and 2002.

The metaverse, sadly, has no Herman Miller to bid on.

You could, however, bid on the plot next to Snoop Dogg’s in The Sandbox. In 2021, at its peak, its price was $450k. It lists for a little over $1k today. Or a 99.8% drawdown.

This piece is not about finding value investment opportunities in the metaverse. Instead, it seeks to understand how the pendulum swings when it comes to the human psyche over the past 30 years through the lens of technology and venture capital. Much like life, the story of markets is one of birth, death, rebirth and the handful of zombies that emerge in between.

We have been writing about crypto for a little over a decade. In that time, both the industry and we have grown old. The technically right term might be “mature”. For the first time in quite a while, we feel the energy shifting. We internalise it by saying different things

• Crypto is fintech now

• It’s just infrastructure, in the backend

• The tokens need revenue! And then the revenue needs to buy the tokens back!

• Maybe it was just a bubble?

The pathway to resolution about our little metaverses, NFTs and WAGMI’s dying has been different for each of us. We have been discussing the shift internally and with a few of our VC peers within the industry. This piece is what came of it.

How can an industry be at the peak of adoption, and also be in a trough of depression when it comes to price? Why is it that we have more people than ever on-chain, and yet founders struggle to raise money? Why is it that crypto is now a “hush-hush” topic that falls considerably below AI, defence, Space, healthcare and perhaps even legaltech?

In 2022, being in crypto may have raised eyebrows because of the fraud in it. In 2026, it would raise alarm out of concern.

Our About page, written a long time ago, clarifies that we write for founders. In pursuit of that mission, this piece is all we could gather about what’s happened to crypto venture as a construct, where it’s headed next, and the parallels in history for the current glut. It is equal parts a memoir and a siren’s call for a future that could be co-constructed.

Let’s dig in.

Before we do, big thanks to Haseeb Qureshi (Dragonfly), Kinjal Shah (Blockchain Capital), Brooke Pollack (Hutt Capital), Richard Chen (Varrock VC), Xavier Meegan (Frachtis VC), Peter Pan (1kx), Michael Cieri (Robot Ventures), Steven Venino (Strobe Ventures), Wally Hansen (Crosslayer), Chad Fowler (Blueyard), Joe Eagan (Anagram), Krishna Sriram (Quantstamp), and Brandon Kumar for conversations that helped shape the article.

We appreciate Fernando Gouveia (Compa Capital), Jack Sun (Wonderstruck), Kunal Goel (LayerZero), Jakub Rusiecki (SGV), Alex Gedevani (Monad Foundation), and Jin Ming Neo (HashKey Capital),Darshan (Polaris) for giving feedback on the article.

Seeding Liquidity

By many measures, crypto is having a landmark year. Institutions hold over $175B of crypto through exchange-traded products. On-chain companies earned $11B in fees over the trailing twelve months. The GENIUS Act is about to wipe out a decade of regulatory limbo. Exits have reached a record, with $8.6B of M&A and eleven IPOs.

These are fertile grounds for the “seed” of venture capital to flourish. But speak to any founder raising in these markets, and they’d quickly clarify that liquidity is nowhere to be found. According to Galaxy Research, only eight new venture funds formed last quarter - the fewest since 2020. Quarterly deployment is down to $4B. Extrapolating annually, that $4B becomes $16B. Or half of what was deployed in 2021 at $31B. While the institutions are validating the asset class, VC dollars that create new categories and assets are yet to catch up.

One heuristic is to think crypto as a sector has gotten so efficient, that it no longer needs as much money to be deployed. The other would be to acknowledge what broke along the way. But to understand what broke, we need to understand how we got here in the first place.

Liquidity was once a privilege given to a firm that has spent long enough building something valuable. The median age of a tech firm listing today is 14 years. During the dot-com days, it compressed to five years. Their valuations also justified sticking around the long haul. In 1980, when Apple listed, its valuation was $1.8B. When Meta (Facebook back then) listed, it had 901M MAUs and revenue of $3.7B the year before. It listed at a valuation of $104B after existing for a little over eight years. ±40% of anyone who interacted with the web, tinkered with Facebook. Why was this the case?

The amount of money allocated towards private market technology investments was a fraction of what it is today. So there used to be a pecking order. A VC would absorb all of the risk in founder selection, sector expansion and management. In exchange, they would get a lower valuation, which could then scale and be exited at the time of an IPO. In 1980, Sequoia had to sell its stake in Apple for $6M in order to produce a 40x return for their backers. They did not have the liberty of being a permanent hold vehicle back then because the asset class was considered risky. Almost like rich-people memecoins.

That lack of belief has played out multiple times over the years. In 1987, the NYT suggested that Pizza shops are a possible competitor for tech investment dollars to flow towards. In 1991, as the first generation of valley tech adopters came of age, Time wrote a piece titled “How Gray Is My Valley.” In 2002, after the dot-com bust, NBC snuck its head out and suggested the asset class had too much capital. By 2013, venture capital evolved into a lifestyle choice, so VCs began warning about funds simply role-playing the job of a venture capitalist. They were dubbed Zombie VCs.

Can’t Eat Pizza Tokens

Crypto’s coming of age happened at the end of decades of tech evolution - with VCs wondering whether they should invest in pizzas and the asset class itself refusing to die.

From 2013 to 2016, there was barely any venture to speak of within the industry. Access to the asset class was through an exchange account, and the asset was the network itself. The first dedicated funds formed around bitcoin; Pantera’s 2013 fund bought at $65.

Launching a token in those years meant launching a chain. A team forked the code, ran the infrastructure, and convinced exchanges to carry the asset. The ICO collapsed capital formation and listing into a single event. Developers wrote their own smart contracts, paid for audits, and negotiated listings with exchanges that could say no. No exchange listing meant no liquidity, and almost certain death.

Selling an idea to the crowd worked as capital formation at first. Ethereum raised $18M this way in 2014 and built all the infrastructure that all EVM chains still rely on. But the crowd eventually got sold just ideas and tokens. Products and infra capable of generating value for these tokens were mostly absent.

This worked well as a narrative. In June of 2017, ICOs raised more than venture capital funding for the industry for the first time. By July that year, ICOs outpaced VC deployment by four times. By December, it was established that ICOs would kill venture capital. (It appears people really do want venture capital dead).

If the primary job of a VC were to simply connect capital to a venture, these arguments could have held weight. But the data tells a different story.

More than half of the projects were dead within 120 days of their sale, and academic work on the period puts the share of outright scams near 80%. $28B was raised between ICOs. We have proved that global-scale capital coordination is possible. But we did not have a mechanism of pricing seed-stage companies with no revenue without guardrails in place. The euphoria had people rushing without questions at first, and with neither profits nor answers later on.

Crypto-VC as a construct emerged in this hubris. Instead of raising directly from a market, teams would raise from a curated subset of capital allocators with a promise of a future token. A SAFE with a token warrant attached became an insulated, sane version of raising. For founders, it meant having the privilege of time on their side as they figured out what exactly had to be tokenised. For VCs, it meant public market liquidity for ventures that struggled to prove private market valuations.

A fund that entered at seed and saw the token list at 4x was made whole by its first 25% unlock, and everything after that was profit. So all a founder had to prove was that they could list the token quickly and manage it reasonably, so that unlocks did not lead to massive sell-offs. This meant that incentives were aligned with the pace of liquidity rather than the value a project might accrue. Fund portfolios became biased towards listings rather than towards assets that represented lasting products.

Investors realised that they could exit before they had to be right. It also allowed them to churn the money faster. Teams realised that when you strip away everything, this is what really moved the needle with the investors. So they did what the investors wanted. They made sure tokens were listed as early as possible. Early token listings took care of funds, and tokens became the primary product. Instead of being a mechanism to fund continued development, tokens were about paying out the cap table.

Mistakes Were Made

Tokens of all kinds, whether governance or utility, failed for one of the two or both of the following reasons:

1. The business model was flawed or non-existent

2. Unlike equity, tokens were not a legal claim on the business

Tokens are a great way to align different stakeholders across geographies toward the same goal. But crypto learned painfully that just because you have a token does not automatically mean you have staying power. Tokens are good to bootstrap a business. Usually, when businesses subsidise something, they are trying to cultivate a new behaviour in their users.

Whether it was Uber subsidising rides or food delivery services making deliveries for free, these businesses used VC dollars to build consumer habits. Once these habits were formed, they could charge users for it to keep making revenue. The habit that businesses were trying to build was directly tied to the business.

With crypto projects, the activity was rarely linked to increased revenue for the business. When X-to-earn emissions ended, the users left with them, because the product had no demand of its own. Much of DePIN repeated the same loop and subsidised supply for demand that never arrived. Suppose you have a token that has successfully bootstrapped a business and has genuine users. But that is still not a sufficient condition. For a token investment to be successful, the token needs to have a legal claim on the business outcome.

Unlike a token, equity is a legal claim on the business. Shareholders can take the board to court when the board’s actions warrant it. Friend.tech’s protocol earned tens of millions in fees while its token captured none of it, because holders had no claim on those economics. So when a token trades below an equity multiple, the discount is often correct.

If the early liquidity and price discovery that tokens enabled can exist through tokenised equity, investors keep the genuinely useful part of the token model and gain real ownership and legal recourse with it.

Tokens have undergone two shifts in response to this.

1. Some have begun swapping back to equity when the market has clearly mispriced what a business is worth. Across Protocol is one instance of this.

2. The breakout winners for 2026 have all tied their revenue back to the token in some form.

In the 1990s, you were valuable by being “on the internet”. In the late 2000s, you could command a premium by being “on a mobile”. In the late 2010s, that trend shifted to being “on a blockchain”. When the tides return, what is left are hard questions about which businesses built in the days of easy liquidity were actually valuable in the days of easy liquidity. Crypto simply seems to be having its moment of reckoning now. Peter from 1kx had a beautiful way of framing this in our conversations.

“Apps didn’t fail because they used token incentives. They failed because they weren’t good businesses. Crypto’s main sin was that there wasn’t enough durable innovation resulting in profitable businesses or protocols with durable product-market fit.” — Peter Pan, Research Partner at 1k(x)

The days of easy liquidity are gone, what is left, then? Have the billions of dollars spent on the industry yielded outcomes? There are sectors that show clear PMF, and the conversation about venture in crypto is not finished until they are mentioned.

Proof Of Work

Three verticals have found durable product-market fit. We have been spending time and talking to teams to make bets on verticals that will find PMF next.

Stablecoins

Supply exceeds $300B, annual transfer volume reached $46T, and roughly $9T of that remains after stripping the bots. Issuers collectively rank as the seventeenth largest holder of US Treasuries. Circle went public and jumped 167% on day one, and consortium banks are now issuing their own. This vertical belongs to growth equity, corporate acquirers, and bank strategy teams now. Issuance now belongs to growth equity, corporate acquirers, and bank strategy teams, and the funds that captured it wrote their cheques when stablecoins were a possibility. The seed window has now moved up a layer, to the companies built on top of the dollars. Ethena was seeded in this vintage and became one of its biggest winners. Bridge went on to get acquired by Stripe within three years.

Prediction markets

ICE committed up to $2B to Polymarket. Robinhood made event contracts its eleventh $100M+ business line. Susquehanna is building a prediction markets business. When the world’s biggest exchange operator, its biggest retail brokerage, and one of its biggest quant firms all arrive within twelve months, price discovery on the vertical itself is complete.

Perp exchanges

Hyperliquid handles roughly 44% of on-chain perp volume and out-earned most public exchanges last year with a team of eleven. Coinbase’s $2.9B Deribit acquisition set the comp for everyone else.

All three now earn billions in annual revenue. They looked like improbable seed rounds four to six years ago. Polymarket was dismissed as a novelty in 2020. Circle spent years as a payments company the market ignored. Hyperliquid launched into a market that treated perps as a solved problem.

Every crypto bear market had some positives. In 2018, the promise was that the infrastructure would improve, that chains would get faster and blockspace cheaper. In 2022, it was that the institutions were coming. Both promises were kept. But the silver lining this time is that nobody has to believe a forecast. These three categories work the way they do because crypto’s architecture enables things that aren’t possible without it. Dollars settle around the clock without a correspondent bank. Exchanges hold client assets in custody that the client can verify.

Markets on anything pool liquidity from anywhere. And the revenue is no longer bought. The largest applications cut token incentives from $2.8B to under $0.1B, and fees still kept growing.

The clearest sign that the categories are real is outsiders trying to build into crypto. Robinhood added event contracts, tokenised stocks, and a chain of its own because they can be the next $100 million revenue lines. Robinhood is doing all of this because building in or adjacent to crypto maximises value for its shareholders.

Crypto was a cool esoteric technology in 2017, but today it’s capable of making money for Robinhood.

Public fintechs build where the margin is, and right now that is here. All of it is happening inside an industry that looks terrible from the outside, where half the tokens listed on major exchanges last year are down more than 80%. The tokens and the businesses have parted ways. A decade of overbuilt infrastructure is finally paying out at the application layer, the same sequence the internet followed when overbuilt fiber made the application era possible.

Venture-scale returns are usually made by betting on verticals that are not yet well established. The next verticals to reach fit will most likely form adjacent to the established ones, because a vertical at scale creates demand at its edges. Stablecoins need credit, brokerage, and treasury products built around the dollars they hold. Perp rails are already extending into equities, commodities, and indices. The job of an early-stage fund deploying in 2026 and 2027 is to back these adjacent verticals before their fit becomes visible in usage and revenue data.

Architect shows how fast an adjacent bet can inflect. It became the first US-regulated derivatives exchange for compute and the AI economy and cleared $1B in volume within months of launch.

But what makes a vintage in 2026 distinctly different from that of 2020 or 2018? A huge part of that conversation is how regulations and mindsets around the industry have evolved. Look beyond the charts and you will realise that the machinery of policy and perception has shifted to be far kinder to the industry than it once was. In the early 2000s, founders building social networks out of the United States had a competitive edge through regulatory protection. A founder (like Zuckerberg) would not be held personally accountable if someone posted things that were deeply offensive to a group of people.

Tim Wu argues in “Who Controls the Internet” that the regulatory environment in the United States is what led to the creation of multiple trillion-dollar companies from the region. Although prices are down, it appears the US has been repeating that playbook with crypto by taking charge of creating frameworks for the industry.

Make Crypto Great Again

GENIUS was signed in July 2025. CLARITY cleared the House with both parties aboard, and the FDIC is writing rules for bank-issued stablecoins. Clearer regulations impact crypto ventures in three direct ways.

1. For a decade, every American crypto term sheet priced in some probability that the government would simply end the company, the way it nearly ended the exchanges in 2023. Every company faced this risk. So, it could not be mitigated or diversified away. The only workaround for investors was to price it into every deal. With US policies shifting in favour, the additional risk premium goes away.

2. The second impact is on founders. According to Electric Capital, the absence of clear US policy may have caused the share of US crypto developers to decline by roughly half over the past decade, to under 20%. Building a crypto company in America is for founders comfortable in grey zones. Government policy changes no longer require founders to navigate legal issues. The same adverse selection that shaped the ICO market in 2017 shaped the founder market for a decade, and it has now reversed in the industry’s favour.

3. The third change is on the cost of doing business. Compliance cost is the last durable moat in fintech. The company that onboards a regulated customer for $40 instead of $400 compounds that advantage every quarter. Licensed crypto rails are machines for collapsing compliance cost, which means crypto companies are being handed the same moat.

Capital usually follows Washington with a lag. When the US led with enforcement, capital formation moved offshore and quality deteriorated, because offshore meant a lack of regulations. US and offshore are both now changing their stance. The UK is writing its own regime, and Coinbase and Robinhood are already building for that market. Hong Kong’s stablecoin ordinance took effect on August 1, 2025. South Korea introduced its Digital Asset Basic Act the same summer, and Japan already regulates stablecoins under its Payment Services Act and keeps refining that framework. Wherever a founder incorporates now, some rulebook applies.

Mixing Revenue

We discussed how the lack of revenue led to the forced listing earlier in this piece. That has now changed.

On-chain protocols earned $11B in the trailing twelve months, and at least a dozen of them earn at a nine-figure annualised rate. Speculation remains the core of that revenue, and crypto’s first fee-paying product was always going to be its casino. But there are degrees to the speculation. At one end, Axiom’s fees track memecoin volumes and fall with them. It collected $100M in fees within four months of launch and roughly $700M since, and its quarterly revenue dropped 86% when memecoin volumes collapsed.

At the other end, Hyperliquid (roughly $843M last year with eleven people) and Aave earn from trading and lending that persist through cycles. Across the market, teams are trying to move their revenue entirely off crypto’s cycle, and Hyperliquid’s HIP-3 markets show how far that has gone. Those markets already carry $4B of its $10B in open interest, with roughly $3B a day trading in assets unrelated to crypto.

Consumer software built on trading primitives is also generating revenue. Phantom did roughly $325M in 2025 from swap fees on seventeen million monthly users, which is app store economics rather than exchange economics. Phantom doesn’t have a token and probably doesn’t need to have one either.

Then come the enterprise sellers. The companies selling tokenised assets and compliance infrastructure to institutions grew the RWA market from $5.5B to $18.6B in a year, and they earn revenue through sales contracts that are independent of broader asset prices.

And furthest from the casino are the stablecoin issuers, whose earnings depend on float and interest rates rather than on crypto prices. Tether reported over $10B in profit for 2025 from its Treasury reserves, and Circle took the same reserve model public. Ten years of crypto produced one Circle. The last two years produced candidates at every degree of the spectrum.

And we are also seeing earnings flowing to investors who hold the asset. Protocols paid $96M to holders in a single month through buybacks, which is one way the token serves as a claim on the business.

A sector view also shows where the market is still adjusting. By late 2025, applications in DeFi and financial services earned 73% of all onchain fees while blockchains earned 12%, yet blockchains held 91% of protocol market capitalisation and the applications roughly 6%. That prices blockchains at nearly 4,000 times their annual fees and the applications at 17 times. The gap has held through 2026. Uniswap, Pump.fun, and Polymarket each collected more fees in the past month than Ethereum or Solana did. The market is still pricing the infrastructure of the last cycle over the businesses of this one, and an investor who knows where the fees are earned can see it.

We spend our time on a handful of sectors, like financial applications powered by blockchain rails, tokenised collateral, and on-chain credit, because we would rather know a few markets deeply than have a view on everything.

The market has started paying for exactly this kind of underwriting. Delphi built a portfolio of the ten largest revenue-generating tokens, weighted by revenue, and tracked it from January 2025 to May 2026. It returned 30.6% while BTC fell 17%, ETH fell 35%, and SOL fell 58%. Tokens with real cash flows beat everything else in the asset class, including the majors. Picking businesses over stories now works in the liquid market, and it works earlier in the private one.

Every asset class that moves on-chain creates a generation of companies around it, and stablecoins have established that template. When dollars wanted to move on-chain, they needed on-ramps, cards, and treasury products. Circle and Tether are a result of that demand. We think something similar is underway for two more asset classes: Treasuries and equities. BlackRock’s money market funds now put Treasury yield on-chain with Securitize as administrator. Tokenised stocks moved ~$23B last month and are growing ~100% month over month.

The pressure behind the equity migration comes from the private markets themselves. The most valuable companies stay private longer, so investors locked inside them sell stakes to one another as they wait. Secondary sales of private-fund stakes reached $240B in 2025, a record, up 48% from the year before. Retail cannot buy these companies at all until they list. Coinbase built the first bridge in June with pre-IPO perpetual futures, starting with SpaceX, synthetic exposure ahead of real issuance. Tokenised equities are the venue where that locked supply and that excluded demand eventually meet.

A third market is forming around machine payments. Payment companies are converging on standards that let AI agents pay in stablecoins, because software that buys things needs money that settles like software. Cloudflare built a payment gateway on the x402 standard it developed with Coinbase, so an agent can pay for an API call or a page crawl in stablecoins without a card network involved.

"Agent-native trust is a completely new category that most AI investors are ignoring. As AI capability compounds, security becomes paramount. We see a clear opportunity in securing the future of multi-agent networks by implementing the past 10 years of research done in blockchain. Crypto-native investors are primed to underwrite what we believe will become one of the largest markets of the next decade." — Xavier Meegan, Founder and CIO at Frachtis

As tokenisation becomes normal, every function around the asset has to work on-chain too. A tokenised asset becomes productive only when it can be used as collateral, and credit is where we are already invested.

Banking is the industry where this migration has gone furthest. Banks cannot innovate at the speed their customers move, so fintech built neobanks. Neobanks reached the limits of the rails they rented, so builders made crypto neobanks. Now come crypto-native banks with actual charters (Erebor received its national bank charter this year) and FX settlement built on stablecoins (OpenFX raised $94M for exactly this). Each generation moved one layer deeper into the stack.

Crypto has stopped being a parallel financial system waiting for permission. It is becoming the plumbing of the existing one, and the venture opportunity of the next cycle is in that merger.

The dollar migration is already far enough along that its users stop noticing blockchain elements. Ramp’s finance customers pay vendors in São Paulo the way they pay vendors in Seattle, settle over weekends when every traditional rail is closed, and barely register that a stablecoin was involved. Visa processed $3.7B in payments through stablecoin-linked cards in the past year, on 1.9 million cards across more than 200 markets, with plans to expand to over 100 countries with Bridge. $9T a year settles this way after stripping the bots.

The crypto part turned invisible, and the companies that made it invisible earned the fees. Equities and Treasuries are at the start of the same curve. So the venture question has narrowed from whether assets migrate to which products each migration needs. Stocks on-chain need brokerages, collateralised lending, and market makers. Tokenised Treasuries need custody and distribution. Almost none of those companies exist yet, and companies like that get built at seed.

Taking The Right Exit, After Tokens

Blockspace is now oversupplied and earns little, while fees have shifted to applications. Hyperliquid earns more than the chains it was supposed to depend on. In 2018, it was one exchange protocol and a stack of hackathon demos. Apps are where the value has been accruing, and they are usually starved of capital. With fintech and crypto fusing, there will be an ever-growing long tail of products that bring these assets to people who will never know what a bridge is. DeFi was crypto’s only financial product line for a decade. Neobanking, brokerage, FX, and credit are each now being rebuilt with crypto built in.

Overbuilt infrastructure allows outsiders to build for specific problems. Vinod Khosla has argued for years that the companies which transform an industry rarely come from its experts. Tesla and SpaceX were built by Elon Musk, a software entrepreneur. Impossible Foods was built by Pat Brown, a Stanford biochemist. Experts think in an industry’s constraints; outsiders arrive with a problem and treat the industry as a tool.

For crypto’s first decade, the tooling was not ready, so building here required believers willing to construct exchange rails, custody, and compliance from scratch. A decade of overinvestment in blockspace and tooling ended that. And with it, the “shape” of what we considered “crypto founders” has shifted too.

• Hyperliquid was built by Jeff Yan, a high-frequency trader from Hudson River Trading who wanted a better derivatives exchange

• . Circle was built by Jeremy Allaire, who had taken two internet software companies public, Allaire and Brightcove, before he touched a blockchain.

• Erebor, the first bank to receive a national crypto charter, was founded by Palmer Luckey, the defence-tech entrepreneur behind Anduril.

Each of them came to crypto with a business that needed rails rather than a belief in search of a use, found the rails ready, and built on them. The idealistic, utopian days of what used to be considered “crypto” are behind. Operators now look to blockchains as just infrastructure. Like Linux. Or databases. The infrastructure itself becomes a commoditised bet that might mimic Cisco from the early 2000s. This is where venture capital as a construct becomes relevant.

For a decade, the token was the only exit, so every company was forced toward a listing whether its business was ready or not, and the failures catalogued above mostly trace back to that pressure. The token remains an exit, and a stronger one now that buybacks and revenue make it a claim on the business.

It is no longer the only one.

M&A set a record at $8.6B with over 130 venture-backed companies acquired, and the buyers were strategic rather than financial. Naver paid $10.3B for Dunamu. Coinbase bought Deribit for $2.9B, then bought Echo. IPOs work again. A crypto company built today can be acquired, go public, or return cash through its token. No earlier point in the industry’s history offered all three.

A longer path to liquidity restores the incentive to build businesses worth holding, and it filters the capital base to investors who can wait. The easy route, launch a token and get immediate liquidity, is closed. This is probably the best thing to happen to innovation in crypto. Founders now differentiate on product and adoption rather than on marketing and emissions, which is what building looked like before 2017, when nobody was here for quick money.

One place we are seeing liquidity is through secondary buy-outs. Brooke Pollack from Hutt Capital used to invest directly into VCs in the past. Currently the fund is looking towards secondaries, both due to the opportunity they represent and the need the asset class has for liquidity within the industry.

At least three quarters of our capital now is buying out LPs in existing funds…so we are now a secondaries fund, though still making some primary allocations within this strategy

As the industry evolves, the pathway to exits would look quite different from what we are used to. This does not mean tokens as a construct are dead. Some of the best names in crypto are powered by tokens. But it does mean the age of easy money from tokens is behind us.

How Do Investors Adapt?

History is a cruel poet with a passion for rhyming. Maybe history is just misunderstood, because in rhyming, it leaves clues about what to expect as time passes.

Deep in the bear market of 2003, Fred Wilson from AVC wrote this note on How Much is Too Much. It is a retroactive look at what happened during the dot-com bubble.

1. There were too many VCs getting funded that weren’t actually VCs. There were people who decided they wanted to be VCs, were able to raise money because there was so much out there, and they set up shop and did a bunch of stupid things. If all of that money went only into the hands of experienced VCs, a lot less of it would have been invested, and a lot less of it would have been lost.

2. The equity markets tanked in 2000. It takes 3-5 years to get a company to the stage where it can be sold or taken public. The health of the equity markets at that time will have a lot to do with the returns that can be generated by a VC. So you had to make investments in the 1993-1997 timeframe if you wanted to take advantage of the crazy prices that were being offered in the 1999-2000 time frame. The “good years” benefited from that. The “bad years” suffered.

One of the arguments being made was that VC as an asset class cannot absorb more than $10B a year. Years later, he wrote a potential fix for the situation. It was to go deep where others went wide. There were no more “internet investors”. Just like perhaps today there are no longer “crypto investors”.

In 2009, he drew a contrast between the two worlds:

Thematic investing involves identifying big themes and going after them. Examples from the world of web services would be “social networking”, “online video”, “ad networks”, “social media”, “real time”, “mobile”. I know many VCs who go about it this way. They identify the themes and then get busy filling out their portfolio with companies that fit those themes.

Thesis-driven investing involves drawing a picture of where your particular area of focus is going. I like to take a five to ten year view. And once you have mapped out that picture, it becomes your thesis. And you evaluate every investment you make in the context of that thesis.

Nine years later, Fred explained one more element that defined how venture evolved.
It was about Hold periods

If you assume the average hold period for an early-stage investment is seven years and if you make one to two investments per year, you will have between seven and fourteen portfolio companies to manage at any one time.

The low end of that range is quite manageable. The high end of that range is not. I have been there.

I believe that early-stage venture capital done right is a service business in which the entrepreneur and the company they started is our customer… Doing this well is hard. Because if you only make eight to ten new investments per year and expect to produce at least one billion-plus exit each year, something we have been able to do every year for almost ten years now, you have to have a pretty high hit rate on super early-stage investments.

The Deja-vu of going into the mind of an investor who saw the dot-com boom, and the rise of unicorns in the years that followed, is quite real. Crypto’s broad arc from here on is defined by the same three themes. A requirement to look beyond the insularity of what made the industry, going deeper towards sectoral expertise and expanding hold times from a few years to almost a decade.

The founders have already adapted. The token-as-the-product playbook is finished. They will be more careful about whether and how they launch a token, and they will not launch one just so investors can churn their capital. As founders change, investors have to change with them. When a company takes five or more years to exit, the investor has to be right about the business, and being right about a business starts with understanding the market in which it operates.

Longer holds also mean that you can’t spread your capital too thin across too many sectors. As the churn rate declines, investors will have to be more deliberate about the businesses they deploy in. Investors’ edge will likely come from specialising in a vertical. Being a specialist means knowing how a sector makes money, who the buyers are, what good unit economics look like, and which regulations decide who can operate.

Cambridge Associates found that US venture and growth equity funds across the 2001 to 2010 vintages returned 2.2x gross on invested capital with a 23.2% gross IRR for sector specialists, versus 1.9x and 17.5% for generalists. The gap held across consumer, financial services, healthcare, and technology. The study also found that specialists deployed more selectively when valuations were high. Crypto venture never had to learn that discipline because the token listing paid for the lack of it.

The other part of adapting is knowing which stage actually needs you. Capital for growth-stage crypto companies has probably never been as available. a16z raised $4.5B for its largest crypto fund, Paradigm and Haun manage billions more, and TradFi funds now join later rounds. A seed company that discovers good businesses will find plenty of capital waiting for it at Series A and beyond. What the market is missing is investors willing to write the first cheque with an informed view. Seed managers who have done the work become the layer that finds companies for all of that downstream capital.

That leaves one question. If the setup is this favourable, why are only a few funding it?

Several of the largest crypto-native firms have raised non-crypto funds or expanded their mandates beyond the asset class, and roughly half of the troubled mid-sized crypto funds are expected to liquidate or convert by 2028. A firm managing billions or hundreds of millions cannot deploy them through $2M seed cheques, so the largest funds broadened into AI and elsewhere to match their size. The mid-sized funds are leaving because performance no longer allows them to raise capital.

The early stage itself remains open, and its round sizes fit small, specialised funds. This is how capital cycles end in every asset class. Opportunistic capital exits, generalists return to their core mandates, and the funds that remain are those whose LPs re-committed on evidence. Fewer than 20 firms actively write pre-seed and seed cheques. Deals that closed in three weeks now take much longer to perform due diligence. Only eight new funds were raised last quarter. A fund usually deploys over the three years after it closes, so very few investors will actually compete for the seed rounds of 2027 and 2028.

Which leads us back to that ancient question markets have been asking since the time Sequoia invested in Apple - Is venture capital in the sector dead? Yet?

The reality is far from it. Throughout market cycles, the amount of capital and the number of people pursuing opportunity have grown and stagnated. But the best funds tend to be the ones that double down when it gets scariest. Collaborative Fund defined the economic reason for this in simple terms a decade back.

Markets reward being right in non-consensus bets.

Crypto today is a non-consensus bet today. The number of people that are deep enough in the weeds to understand what may be right is few and far between. That is the economic case for new funds being raised within the industry today. Kinjal from Blockchain Capital explains in the words below.

"The best entry points never feel like consensus. That understanding has anchored our conviction in digital assets from the start, knowing this is a story that plays out over decades, not cycles. Today, capital continues to consolidate and adoption is reaching record levels across use cases with proven PMF. We believe digital assets will redefine financial markets and networks, and we are excited to lean in during the time when others are looking elsewhere." — Kinjal Shah, General Partner at Blockchain Capital

Patient Capital

The arc of crypto rhymes with that of the internet. It started out with an ideological pool of users, who gathered without a business model. It then transitioned towards nation-states and governments embracing it, while the initial euphoria of investment dollars and novelty faded out. People often draw parallels with crypto and the dot-com bubble.

But perhaps, we are closer to 2009 than we are to 1999.

The infrastructure is mature, the regulations are clear, the use cases are here, and sentiment is at one of its worst. Deep in the weeds of that crisis, a handful of firms generated some of the best DPIs that the venture asset class has created. History, as we’ve seen repeatedly, is a cruel poet with a passion for rhyming.

Here’s why: The founder pool has thinned to its strongest cohort. As the opportunists left, the average teams left with them, and a small group of exceptional teams remains with very few venture dollars competing for them. LPs no longer need convincing that crypto matters. Billions in ETP money did that for them. What they are choosing now is the manager. Strategies in crypto change with each cycle, so LPs back people who have adapted across cycles rather than relying on a single playbook. Specialist knowledge also wins deals.

Founders pick investors who understand their business before the first call, and that knowledge takes years to build within one ecosystem.

When the money does return, it may not even be called crypto investing. Nobody calls themselves an internet investor anymore. The internet mattered so much that the category dissolved into several sectors or verticals. Amazon was an internet company in 1997 and is a retailer today. Netflix was an internet company and is a studio today. The investors who profited backed them while the internet was still treated as a niche, as Kleiner Perkins did with its $8M cheque for Amazon in 1996. Crypto is on the same path through finance. The companies in this piece will increasingly be called brokerages, payment companies, and banks, and the funds that back them will simply be called venture funds.

Finance is absorbing crypto one function at a time. Banks are receiving crypto charters, stocks and money market funds trade on-chain, and payment networks are using blockchains that allow programmatic money spends. In aggregate, this is an enormous market. Trust functions like verification, settlement, and safekeeping that every transaction requires account for roughly $29T of global economic activity. AI collapsed the cost of intelligence. Stablecoins are now collapsing the cost of trust the same way. No economy has ever gone back to a slower way of moving money. So any function that migrates to crypto rails doesn’t go back to old ways.

Every migrated function needs a generation of products built around it, and small teams build those products on a few cheques. The cheques also go further than they used to. The capital required to start a company is falling as AI writes more of the software, which shifts the scarce inputs to judgment, distribution, and patience, at exactly the moment the field emptied.

What this new age needs is what VCs have always provided when sentiment is low. In 2009, Fred Wilson defined it as Slow Capital. In 2026, Will Mandis calls it Patient capital. In this new age of generative slop and hyper-financialisation, the ability to sit out patiently for the right opportunity and the “taste” to identify what is right is where all of the value in the next crop of venture funds emerging within crypto lies.

Doubling down,
Saurabh Deshpande

This piece builds on his perspective. I take a look at a handful of startups that are building the orchestration layer for global finance atop blockchain rails. My observation is that blockchain rails will become infrastructure for the export of American institutions and assets around the world. The piece today lays out a thesis for why that could be the case. This is one of a series of notes we plan on releasing on how our internal thesis around crypto and blockchain rails will evolve from here on.

If you are a founder, we’d love to chat. Join our community here.

In the opening chapters of The Splendid Exchange - there is a reference to Herodotus’ account of silent trade between Carthaginian merchants and an unnamed Libyan people on the western coast of Africa. The Carthaginians would arrive by ship, unload their goods on the shore, return to their vessels and make smoke. The locals would then come down, inspect the goods and leave gold beside them. If the Carthaginians felt the price was unfair, they would simply wait for the buyers to leave more gold at the coast. This would repeat until both parties had a mutually agreed upon price. There would be no words exchanged, but there was a shared language.

This was 2500 years back - before World Bank, SWIFT, blockchains and perhaps, most civilisation. All these years later, commerce has been the same. An attempt at exploring shared languages to find fair price. The price, in market terms could be what we consider interest rate, forex costs or the dollar amount to be paid. But all commerce has been an attempt at creating a unified, shared language.

Organisations like the World Bank and the IMF helped create a shared institutional language for sovereign finance. SWIFT did the same for banks communicating across borders. Visa, Mastercard and the DTCC extended that grammar into payments, cards, clearing and settlement. For over 70 years, global commerce has relied on institutions that make strangers legible to one another because part of what maintains global peace is commerce.

If your economy relies on a nation conducting business with you, odds are quite low you would want to bomb them. Intertwined economies, by extension often lead to periods of lower conflict. But these institutions are restricted in their ability to enforce trust. The trust - is codified in legislation, banking relationships and operational processes, but neither verifiable, nor enforceable through a shared technology.

Perhaps a global ledger that is constantly validated by all parties involved could lead to a better solution.

Visa realises this. Mastercard realises this. The DTCC realises this. Even SWIFT realises this. The adoption of blockchains in its different forms is fuelled by a desire to create a shared language for global-scale commerce. We have been operating internally with a simple thesis. The idea is that - blockchains are about to do to capital market assets, what the web did to information. The marginal cost of access will diminish to zero, verification will become faster and the globe will turn into a single capital market.

Tokenisation is an attempt at creating shared languages for the world’s economies to speak to one another. But the assets that move through this language are only as strong as the institutions backing them. An economy is aided by strong institutions that know to balance risk (through interest rates) with balancing the interest of its people (via regulations). America stands out as a nation with strong institutions that allow technology to be developed, exported and capitalised effectively. Silicon Valley may not exist elsewhere, if it weren’t for the capital markets and the nature of institutions in the United States. Delaware is where the world’s startups incorporate because it has access to both capital and the institutions that make these startups possible.

Dollars are a mechanism for the world to benefit from those institutions. Stablecoins, by extension, are an export of the stability American institutions offer its own people. Tokenisation is the grammar that enforces this shared language. The numbers back my claim.

Total stablecoin supply today is at roughly $315B. Tether alone has roughly $141B in direct and indirect exposure to US Treasuries, ranking 17th overall among global holders of US government debt, above South Korea and the UAE. In February 2026, stablecoins settled $7.2T in a single month, surpassing the US Automated Clearing House (ACH) network for the first time. The world’s economy is coming on-chain. Tokenised equities, credit and vaults are how that transition occurs.

As we stated in this post - blockchains allow fintech applications to turn into full-scale platforms.

That means developers from around the world would soon be able to build products to cater to users using the assets offered by these fintech products.

• In Robinhood’s case, that is access to tokenised stocks and real-world assets (RWAs).

• In Centrifuge’s case, it is access to tokenised AAA-rated credit through products like JAAA.

• BlackRock’s BUIDL does the same for money market funds, while

• Ondo is bringing public securities on-chain.

• Apollo and Hamilton Lane are bringing private credit and private market funds on-chain through Securitize.

• Superstate is building for public companies to issue and trade shares on-chain.

• You can even buy access to Blockchain Capital’s tokenised venture fund on-chain.

But users are not onboarded to them directly. Dollars are the hook for most emerging markets. Exporters, freelancers - practically anyone in an emerging market earning in dollars quickly realise they have an instrument that has three functions - all desirable

1. the asset is inflation proof compared to local currencies. simply holding dollars translates to more capital over time

2. the asset can speak to global capital markets through hyperliquid, tokenised stocks and meme assets

3. it is also highly mobile and can be sent anywhere in the world at the click of a button.

In many emerging markets these dollar representations can trade at a premium. In stressed markets and specific periods, that premium has reached 10-15%. The businesses that are able to speak to flows from regional economies and translate it to global ledgers will prove to be tremendously valuable. These businesses take context from regional institutions and make it legible to the globe.

It is a repeat of the web. Google Maps mapped out the world’s streets. Your social feed mapped out its culture. Now a new crop of products map out where assets originate and flow.

We are witnessing several startups tackle this in unique ways.

Key To The Rail

Export and import is the most frequent form of communication between nation states. Swift is the messaging layer atop which that communication occurs. But depending on the region, the pace at which transactions clear can widely vary. If you are an importer in Nigeria, buying from China, your vendor may demand 60% of the payment upfront in dollars.

But the dollars are often not readily available. Nation states restrict access to dollars to maintain their own currency reserves. A USD wire to China - assuming you have the right relationships, at the right banks, with the right amount in bank balances, can take anywhere between 7 to 10 days. And in most cases, people do not have access to all three.

If a vendor in China takes the payment in stablecoins, they forego tax rebates worth up to 13%. So you have two economies where people do want to talk to one another in trade, but the rails do not facilitate it. Part of the reason is documentation challenges, the other is risks with fraud and the last is the challenges in litigating across economies if something goes wrong.

Keyrails comes in less as a lender and more as a payments-and-credit orchestration layer for such transactions. It does not lend from its own balance sheet. Instead, it acts as a clearing house that connects importers with external capital providers - non-bank financial institutions (NBFIs), fintechs and, more recently, on-chain vaults - while controlling the payment path through which the borrowed capital can be used. Lenders earn roughly 15-20% APR, while borrowers pay 20-25%. Keyrails keeps the spread of around 2.5-5%, alongside fees from the payment rail.

Keep in mind, this is the borrow rate, not the conversion rate, which is the parallel alternative for most importers.

Keyrails absorbs Naira in the form of USDT, once it is converted at a regional OTC desk or exchange. Keyrails then forwards payment to the supplier through its SWIFT rail. Lenders pull three months of trading history via API and underwrite the loans in around three hours. The money never goes to the borrower. Instead, capital is matched directly against the seller’s invoice and settled in China through SWIFT.

The math works because the borrower is comparing Keyrails against the parallel-market FX premium, not against a cheap bank loan that does not exist. If, during stressed periods, the status quo costs 20-30% in currency conversion premium and takes 7-10 days to clear, a 20-25% APR facility that settles in six to eight hours can still be cheaper, especially when the loan tenor is short. At a three-month tenor, a 20-25% annualised rate implies roughly 1% interest for the period, before fees and collateral effects. That is materially lower than paying a 20-30% FX premium upfront.

For the seller, the benefit is equally straightforward: dollars arrive directly in their bank account as a compliant named payment, preserving access to local tax rebates that stablecoin settlement would not provide. Part of the reason is the lack of liquidity around these currency pairs in emerging markets.

In this case, Keyrails’ primary value is not that it has the cheapest capital. Its value is that it controls the rail that makes the lending safe. It standardises how trade data is gathered digitally, creates payment rails across countries, and restricts capital within its ecosystem so borrowed funds can only be consumed against real invoices. Its moat is the combination of underwriting data, compliant settlement and use-restricted capital. With each transaction, the system becomes better at making regional trade flows legible to global dollar capital.

Programming Credit

We noticed one of our own portfolio companies focusing on a different part of the equation. Keyrails focuses on making money move faster between unorganised sectors. Semiliquid provides infrastructure for collateral to stay stagnant, between known counterparties that share collateral.
The shared idea is containment. In Keyrails’ case, funds can only leave through a permitted payment rail. In SemiLiquid’s case, the asset does not need to leave custody at all.

Tokenisation by itself only changes how an asset is represented. It does not automatically make the asset useful as collateral. If a bank, fund or trading desk holds tokenised Treasuries, money-market funds, equities or credit instruments with a custodian, the asset becomes far more productive if it can be borrowed against. Without margin, a tokenised asset is mostly a digital wrapper around something that already existed. With credit, it becomes part of a larger financial machine.

SemiLiquid’s Programmable Credit Protocol (PCP) allows borrowers and lenders to finance tokenised instruments without moving them out of custody. The borrower keeps the asset with the custodian and can continue earning the underlying yield. The lender receives an enforceable claim over that asset. If the borrower repays, the lock is released. If the borrower defaults, the lender can take control of the collateral according to pre-agreed rules. The useful phrase here is delivery-versus-lock: cash can move, but collateral stays where it is until it needs to be enforced.

Consider the math at play. Assume an institution holds $100M of tokenised treasuries yielding 5% annually. If it borrows at 98% loan-to-value (LTV), it can access $98M of liquidity without selling the underlying asset. At a 6% annual borrowing rate, the headline interest cost on the loan is $5.88M per year. But if the Treasuries continue earning 5%, the collateral generates $5M in annual yield. The borrower’s net cost is therefore roughly $880K, or about 0.9% of the $98M borrowed, before protocol fees, haircuts and any custodian fees.

The borrower is not simply paying 6% to borrow. They are paying the spread between the cost of debt and the yield retained on the collateral. In a traditional setup, that yield may be eaten by the bank or custodian. With programmable collateral, the asset can remain locked for the lender while the yield still flows back to the borrower. SemiLiquid is bringing this logic to a broader network of tokenised assets and institutions that want to lend and borrow against one another.

The difference is in the assets involved, the speed of underwriting, and who gets to keep the yield.

This is different from a DeFi lending market like Aave. Institutions do not have to move assets into an open smart-contract pool, accept public liquidation mechanics, or price in the risk premiums that come with permissionless lending. They can keep assets with regulated custodians while still making those assets financeable. For lenders, the upside is faster diligence and cleaner enforcement. The custodian can attest to the state of the collateral in real time, while the protocol prevents the same asset from being pledged twice inside the same rail.

One way to understand why this matters is through collapses like Three Arrows Capital and Archegos. 3AC left creditors with roughly $3.5B in claim. Archegos created a similar blind spot in traditional finance: Bill Hwang’s family office built overlapping swap exposures across multiple prime brokers, and Credit Suisse alone lost about $5.5B when the unwind came. In each case, the problem was not just that prices fell. It was that lenders did not have a shared, real-time view of what collateral existed, where it was pledged, and whether the same balance-sheet strength had been represented to multiple counterparties.

SemiLiquid tries to replace that trust gap with real-time collateral verification. The loan does not need to be publicly visible to the market, but the relevant parties can know whether the collateral exists, whether it is locked, and whether it can be enforced.

Traditional secured lending often requires lawyers, back offices, custodians and manual reconciliation to coordinate around a single financing transaction. SemiLiquid compresses that process into programmable credit infrastructure. More importantly, it turns idle tokenised assets into collateral that can support borrowing, margin and repo-like activity. That matters because the next phase of tokenisation will not be about putting assets on-chain. It will be about making those assets useful enough that institutions can earn a few extra basis points, borrow against them, and use them without giving up custody.

Eating the $Chip

GPUs are the heart of the AI economy. They are what compute the answer when you prompt endlessly on a Thursday afternoon. They also cost a lot, which is why Meta, Amazon, Alphabet and Microsoft spent roughly $410B on capex in 2025 and plan to spend roughly $725B in 2026, an increase of around 77%.

In simple words, GPUs are what tractors would have been in a world where agriculture was the primary source of income: productive machines that convert capital expenditure into recurring output.

Firms take lines of credit to buy GPUs on the assumption that the hardware will generate revenue from third parties leasing compute. But data centres and smaller AI infrastructure companies often do not have access to the same low-cost capital as Google, Amazon or Microsoft. It is an underserved part of the market where even calculating loan terms is difficult, because the price, cost-effectiveness and residual value of GPUs can change as models become more efficient and enterprise demand evolves.

As of writing, USD.AI has roughly $398M in total value locked (TVL), with about $202M deployed in active loans. Loan sizes have grown from $1-5M at launch to a $98M facility against a 2,304-GPU cluster - the collateral model scaling well past pilot size.

Lenders on USD.AI get access to a market driven by hyperscaler demand and AI compute growth. Borrowers on the other hand benefit from financialising an existing asset in hand, without needing to find traditional lines of capital. All while having faster access to money. Both parties benefit from the fact that the GPUs can be financialised in the form of collateral in the event that something breaks.

USD.AI’s core customer is not the data centre alone. Or a startup. Or a hyperscaler. It is anyone interacting with the GPU economy. Its value is in creating a liquid lending market around an asset that is productive, physical and hard to finance. And that underwriting takes years of expertise because a single bad loan can question the legitimacy of the entire market. In other words, they are taking GPU-native context and translating it to a global liquidity pool that seeks to lend against it using stablecoins.

Context is Where the Dollars Are

Each of these businesses look very different from what we used to consider “DeFi”. They use tokenisation, blockchain rails and stablecoins to accelerate markets that were traditionally underserved or disorganised. The moats in these products do not come from code alone, but from deep, enriched context that comes from each transaction cycle on the product. A loan on SemiLiquid’s PCP, a borrow on USD.AI’s CHIP and a transaction on Keyrails are similar in that, with every transaction, the platform generates trust and credibility that is hard to replicate. It also gives these businesses a way to double-check their own processes.

The value in these businesses will not come from transaction volumes alone, but from the context they build on the users, borrowers and counterparties of their products. This is very similar to how banks work. A bank can derive more value from each marginal customer in the form of a credit card, mortgage or fixed income product the longer the customer stays with the bank. In turn, these businesses manage to escape the cyclical seasonality that comes with crypto. Unlike exchanges or trading products, the demand for export-import, loans and margin is steady and applicable across the economy.

In other words, this is how crypto grows beyond trading and speculation into the operating system for real-world capital formation.

Commerce needed a shared language. Tokenisation gave us words that are accessible, easily relayed and transferred. Protocols give us the grammar to enforce it. Startups that contextualise and onboard the world’s markets to this shared culture of transaction will capture the next decade of value.

Exploring Token Aggregators,
Joel John

On May 17, 1792, a group of twenty-four brokers and merchants rallied around a buttonwood tree on Wall Street, New York. Together they agreed that 0.25% would be the absolute minimum that would be charged on a stock trade. By 1975, between tiered costs and complex pricing, that number would inflate as high as 1 to 2%. In 1975, the SEC would deregulate the NYSE’s pricing - setting the stage for what we now know as zero-fee exchanges.

When a transaction becomes as easy as a click, what do you even charge for as an exchange? Robinhood and Charles Schwab now run exchanges with zero commission. Lighter has been teased with bringing the model to crypto.

The story of financial exchanges is that of costs gradually trending to zero. All transactions are simply the movement of information about who gives what asset (usually dollars) in exchange for another asset. If the cost of storing, moving and validating information has gone lower due to Moore’s law, what would be the fate of exchanges? Do they have more value to be created by being a full-stack banking product? Are there parallels in traditional markets?

This is what I set out to find answers to in today’s piece. It is a sponsored collaborative exploration written with the team at Grvt. We have been obsessively studying perpetual exchanges for the past quarter at DCo. We did well taking a position in Hyperliquid last year. Our cheque into Drift followed in Feb.

Grvt’s name repeatedly emerged in all of the charts we studied. What made our interactions refreshing was that instead of looking at exchanges as platforms for speculation, their philosophy was to build the whole stack for a transparent, verifiable brokerage. How it works and why it matters is what we will explore today.

Let us begin with an understanding of how exchanges became what they are.

Trending to Zero

Right around the time we were on our mission to the moon, in 1969, Wall Street was running a different experiment. They began asking what if we allowed large institutions to trade through a network digitally instead of having to deal with the human operators at the New York Stock Exchange? Surely, they would have less liquidity as the network would have fewer participants, but they would have instant accessibility to one another. The system was called Instinet. Created by Institutional Networks Corporation in 1969, the exchange was created for institutional investors to buy and sell orders to one another through a computer network. It was an early instance of what would eventually be known as an Electronic Communication Network.

The NYSE had its own electronic trading network as early as 1984 - named SuperDOT. It too focused on allowing orders to be taken digitally. In 1987, when the Black Monday crash happened, the value of the network became apparent. Traditional avenues that took orders through telephone lines were saturated as investors rushed to sell. ECNs like Instinet continued to allow large market participants to trade with one another. By 2001, Instinet would be responsible for close to 15% of all Nasdaq volume and a considerable share of NYSE-listed stocks; SuperDOT was doing close to 99% of NYSE’s volume.

Both these networks focused on institutions. The next evolution would focus on allowing everyone to trade.

When Charles Schwab launched online in 1996, online trading still represented only a small portion of brokerage activity. Within three years, it had completely transformed the industry. By 1999, Charles Schwab had more than 3.6 million online brokerage accounts, and the firm was processing over 1 million online trades per day through its internet platform. In the early 1990s, a typical retail stock trade cost between $40 and $50 in commissions. By the end of the decade, many online brokerages were offering trades for less than $15.

Financial markets had undergone a two-step transformation.

1. The 1980s were about the gradual digitisation of markets.

2. The 1990s were about its democratisation.

Thanks to the internet and improvements in computer hardware, everyone could trade from the comfort of their homes. The dot-com boom became the perfect macroeconomic environment for these trends to converge and bring trading online.

Over the next two decades, that $15 sum would also trend to zero. In the 1970s, your stockbroker could charge up to 2% because there was a complex network of paperwork behind the scenes each time you purchased shares of GEICO alongside Warren Buffett. The stock settlement would take five days; the reconciliation from your bank to the stock custody avenue would take time. All of this required human labour. What do you charge for when a stock transaction is just the click of a button? How do you justify charging to click that button when there are multiple peers like Robinhood, Interactive Brokers and their regional variants willing to do it for free?

This would be the billion-dollar question that sets the stage for zero-fee brokerage platforms. As early as 2001, the Bank for International Settlements teased the idea that Moore’s law would change how market pricing works. If everyone can have powerful computers, and networks can move increasingly larger sums of information, then it will change how we transact. That transition did happen. By the late 2010s, high-frequency trading and algorithmic orders were responsible for bringing more liquidity and efficiency to the market. Robinhood and Charles Schwab made zero-fee trade structures common by 2020.

How do they make money then? The BIS, in 2001, speculated that exchanges would make money through payment for order flow and have their own market-making divisions. In the 1980s, an order would cost you money. Today, in aggregate, that order should have you being compensated instead because exchanges bundle that information in aggregate and sell it to firms like Citadel. The exchanges have marginal costs in clearing a transaction. Their earnings, however, come from selling aggregate information of how their users are placing orders. There would be one more avenue for these exchanges to make money. That would be sweeping accounts.

When you move $10k to Robinhood to trade crude oil, the money does not sit idle. Until you take a position, it can be routed by Robinhood into a high-yield account paying Robinhood two to three per cent a year. The $200 on your account alone may not seem like much, but in the aggregate it becomes valuable. Robinhood had close to $32.1B in cash-sweep balances in FY 2025. Charles Schwab owns a bank, so it can take its own customers’ deposits, put them in its own bank, and keep a larger share of the interest rate. (This post from 2019 is a great breakdown of how discount brokerage’s unit economics work.)

Both these businesses are economies-of-scale ventures. These things work only when there are hundreds of billions of dollars in idle deposits. For scale, consider that Robinhood had $324 billion in combined platform assets. Of which, $32.4B was just cash. Crypto-native perpetual exchanges, in contrast, have peaked at $5B in cumulative TVL. This is why most exchanges still have transaction fees. Both Hyperliquid and Lighter have explored directly receiving a portion of interest rates generated from USDC deposited on their platforms. But at the heart of all challenges with perpetual exchange monetisation within crypto is the fact that we have not yet evolved to be an economy-of-scale ecosystem. To have economies of scale, we need tooling that grows wealth without requiring individuals to monitor the situation constantly.

To understand where this goes, we need to understand how the perpetual exchange ecosystem within crypto itself came to be.

So Decentralised

Decentralised perpetual exchanges are the third-highest revenue-generating category within crypto. The largest is stablecoin issuers (like Tether and Stable), with decentralised exchanges (like Uniswap) a close second. Both these categories have benefitted tremendously from the network effects of having over eight years to evolve to be what they are. Perpetual exchanges, in contrast, have been a recent phenomenon.

For context, cumulative volume on all perpetual exchanges each month was at $62B in Jan of 2022. In Jan of 2026, that number was at $964B. The same month in 2022 saw only $100k generated in fees across the category. In 2022, it had $3.3M in fees generated in a single month. Cumulative volume across all exchanges hit $14 trillion as of writing this note. Of that, close to $4T was in Q4 of 2025 alone.

Perpetual exchanges as an ecosystem produce the king of hockey stick charts that have conventionally been witnessed only in Web2 software products.

The reasoning for it is fairly straightforward.

1. First, perpetual exchanges have now reached a level of maturity where the experience is at parity with centralised exchanges. Users would rather hold on to their assets and have immediate withdrawal than take on the risks of a centralised avenue. Unlike FTX, asset custody risks on these exchanges are fairly low.

2. Second, they allow users to have exposure to new assets far faster than centralised avenues like Coinbase. In Q1 of 2024, Hyperliquid was consistently earlier than Binance to list new meme-economy assets. The trend continued in Q2 of 2025, when the same exchange began listing pre-market perps for newly released tokens. When SpaceX and CRBRS were listed earlier this year, perpetual exchanges were the best avenues to trade them pre-listing.

3. Perpetual exchanges can generate far more monetary velocity than peers like lending. In the process, they generate fees. For instance, the average day in Jan of 2026 saw each dollar of TVL on Hyperliquid’s HLP vault generate $24 in volume. Last month, each dollar in TVL on Hyperliquid alone generated $0.0053 in fees.

These exchanges are primitives that combine the best aspects of crypto in a single market. They allow global-scale price discovery, real-time settlement, and instantaneous movement of money. Users do not have to worry about where their capital resides, as it is often verifiable via smart contracts. Perpetual exchanges also vastly expand the surface area of what can be traded upon.

Startup equity, commodities, and indices can all find an avenue to be traded and settled via perpetual exchanges.

Perpetual exchanges help take the target addressable market for all of crypto from a small community of spot traders who often traded crypto-native assets to the world. Most of the assets traded in the first wave of spot-native decentralised exchanges were oriented towards individuals from within the industry who knew what tokens were used for. It could have represented a governance asset, a utility token or simply memetic value. But the vast majority of individuals on the web did not care enough about them. By expanding the asset types, perpetual exchanges made crypto-native financial infrastructure relevant for the whole world.Instead of having users trade niche assets, perpetual exchanges create global-scale markets for commodities and equities.

Annual volume has evolved from $648B in 2023 to $6.7T in the past year.

As with any sector expanding that rapidly, perpetual exchanges have seen competition emerge. When multiple players are involved, part of the focus shifts towards business models. We saw Lighter mount a challenge for a brief while with its low-cost fee model, but that has not quite stood up against the behemoth. Part of the thesis is that the most productive aspect of an exchange is the yield it can generate, since exchanges naturally become hubs for individuals to park assets as collateral.

Who owns the yield?

If you were a banker in the 15th century, you would see a golden age for finance. Double-entry accounting was starting to pick up steam in Florence. Foreign currency exchange and lending across cities were also coming of age as trade between medieval-age cities in Europe picked up steam. Part of what was up for debate was lending - and lending rates. Interest rates have long been a heated matter of discussion within finance. Shakespeare’s work famously involves a creditor demanding a “pound” of flesh from a borrower who defaults on his debt. Usury - or the act of charging excessive amounts of interest - has been considered deeply sinful in human society for quite some time.

As humanity evolved and trade went global, longer lines of credit became essential. Against the backdrop of the manias of the 18th century, where Newton famously lost his money, was the evolution of lending markets and insurance. Many of these ideas are caught well in The Price of Time if you would like to read it. Perpetual exchanges have unique interest rate markets in that the “productive” use to which the capital is put happens within the ecosystem of the platform itself.

In conventional lending, like the one where Goldfinch unfortunately lost money by underwriting loans to real-life businesses in Kenya, the money moves out of the system into bank accounts that are off-chain. The “real world” accounts for such risks through requiring strong collateral and creditworthiness checks by multiple players. In fact, private credit is having a bad year: despite the checks and balances put in place by traditional players like Apollo or Blue Owl, the risks of lending to software firms in the age of AI could not be accounted for.

In a perpetual exchange, such forms of lending are in a closed ecosystem. That is - capital rarely leaves the platform itself. When you go “long” on an asset like SpaceX, the platform does not hold the underlying. Instead, what you have is a synthetic instrument that is presumed to be following the price of the real asset in stock markets. Exchanges pull an index of prices from external venues to set a “mark” price - the reference for the underlying that decides whether a position should be liquidated. Nasdaq now feeds price data directly to oracles like Pyth - which is how that price mechanism itself comes on-chain.

If a platform has too many people crowding into a long position on a hot asset, like a pre-IPO stock that is highly in demand, the market incentivises people who are willing to go in the opposite direction. That is, to go short. This incentive, or correction mechanism, is the funding rate. It is the amount of interest a person pays a counterparty for the position they hold - charged by those that are long an asset to those who are short, or vice versa, depending on how demand for the asset sits. In such systems, the interest rate is charged and settled within the ecosystem itself.

Binance, for instance, currently pays a funding rate that can be annualised to roughly 11%. For niche assets with high demand, that figure can skyrocket to over a few hundred percentage points annualised.

The funding rate and open interest are entirely different concepts. The funding rate is the interest rate a person would pay on their position. The open interest, on the other hand, is the size of positions cumulatively held by all the traders on the platform. On Hyperliquid, for instance, roughly $5 billion of USDC posted as collateral currently supports over $10 billion in open interest - the exposure runs to roughly twice the capital sitting behind it. Now why am I mentioning this?

Because these aspects help make perpetual exchanges interesting from the perspective of yield.

1. Unlike most vehicles for lending throughout history, perpetual exchange interest rates come from a closed loop of participants within a platform. Decentralised perpetual exchanges take this one step further by making it easier to verify net platform balance. The risks involved are therefore far lower.

2. Blockchains make the movement, settlement and claiming of interest generated from such avenues far easier than traditional ones. Users can have their interest rate settled directly in a matter of minutes.

There is one more angle at play here, and that is platform deposits. When a user deposits margin at an exchange, the underlying is idle capital. The assumption is that the user may not need to access it immediately, and therefore it can be put to more productive use. Remember how Robinhood used to sweep its user accounts into better sources of yield? Crypto has a version of this. Binance’s BUSD once did a version of this - Paxos issued the coin and parked the reserves in Treasuries, then split the interest earned on them with Binance. Regulators in New York had them shut it down in 2023. Binance could no longer earn the float interest.

In 2025, Hyperliquid found itself in a similar position. USDH, issued by Native Markets, was a stablecoin on the exchange. Roughly 2% of the $5 billion held on the exchange was in the form of USDH. When conditional markets were to launch (HIP-4), the payouts were to be in USDH. The reason for supporting USDH was that Native Markets was willing to allocate 50% of its reserve yield towards buying back HYPE from the market. Instead of focusing on float income, the token was trying to find distribution and value accrual to the underlying. In May, Coinbase acquired the rights to the USDH brand from Native Markets. They also became the official USDC deployer on the exchange. The yield on the $5B of USDC is now being routed back to the protocol to fund HYPE buybacks.

This capture of “float” income is not new in the exchange business. As of last year, IBKR earned more than it did on its trading commissions from its interest and margin business. Close to 57% of its revenue - $3.56B - was from interest as a line item. Robinhood earns close to 34%. Charles Schwab also has about 27% of its revenue coming from interest on sweep accounts. Morgan Stanley takes this one step further by allowing its users to have a wealth management product within its product line. They acquired E*Trade to build further on their interest rate item.

Crypto-native exchanges struggle to monetise the float efficiently. When it does, the token is where that value accrues today. The average user who wants to protect their wealth may not be interested in income generated going towards token buybacks. Today, some exchanges allow users to hold dollars on their exchange balance and receive a yield. However, it often cannot be used as collateral. There is no mechanism for a user to deposit dollars into a fixed income product, use it as collateral and have instant liquidity on it. In traditional markets, such products involving credit, structured derivatives and fixed income do exist. They can be used as collateral, but they require minimum account balances, long onboarding times, and restrictive asset bases to trade on.

Ultimately, for crypto-native derivatives markets to evolve past where they are, users will need three things.

1. Asset flexibility: the ability to switch between crypto, equities, commodities, and yield-bearing products without leaving the platform.

2. Instant liquidity: the ability to access capital when an emergency arises, instead of being locked into long-dated structured products.

3. Interest-rate income: the ability to earn yield on idle collateral while still being able to use that capital productively.

A user could choose a traditional alternative like IBKR if they think the mix of assets there is more varied. A conservative user may come primarily for yield and trade only occasionally. Liquid structured products let users access capital when needed without fully exiting their yield strategy. Fortunately, crypto infrastructure has reached a point where it can do just that.

Grvt uses ZKsync and a mix of crypto-native primitives to address all three issues. Strap in for a bit because I will have to explain the technical choices that help Grvt offer what it does before we look at why it matters for the end user.

The Pull of Grvt

When Saurabh wrote this note on ZKsync, he clarified that Prividium allows traditional financial players to build decentralised products without giving up control over how assets settle to mainnet. Financial institutions need to balance privacy around user transactions and balances with the ability to intervene when something goes wrong.

Grvt is a useful example of that design philosophy applied to markets: private execution, user control, and Ethereum-linked settlement. It is one of the leading apps within the ZKsync ecosystem. Each day, it processes a little over a billion dollars in volume. As of writing, it has roughly $46 million in TVL and $348 million in open interest. Among perpetual exchanges, it ranks fifth by 24-hour trading volume on DefiLlama’s perps ranking, behind Aster and edgeX.

It can deliver on that because it has abstracted most of the complexity behind decentralised exchanges. I used the product for a few weeks over the course of writing this story. You sign in with Google, set up 2FA and manage margin variables the same way you would on Binance. The experience is heavily mobile-first, as Grvt expects retail users to come through mobile apps first. Part of the reason it focuses on retail adoption and scale is visible in who sits on its cap table.

As of writing, Grvt has close to 90,000 users. In September 2025, the firm raised a $19 million Series A led by ZKsync, with participation from Further Ventures, EigenCloud and 500 Startups. Across all rounds, Grvt has raised roughly $33.3 million. In other words, Grvt has grown into one of the more visible new perpetual exchanges of 2026 while still having enough capital behind it to keep pushing further.

Notice that ZKsync led the round. There are a few reasons for it.

First, ZKsync gives Grvt transaction throughput at scale. Grvt claims it can process close to 600,000 transactions per second, or 36 million transactions per minute. For context, Hyperliquid supports roughly 200,000 orders per second, while Binance’s matching engine is stated to handle 1.4 million orders per second. The units are not perfectly interchangeable, but directionally, Grvt’s claimed throughput is above today’s leading on-chain perp venue and roughly 43% of Binance’s stated matching-engine capacity.

The second part is ZKsync’s use of Atlas. Atlas is ZKsync’s interoperability and performance layer, designed to let ZK Chains communicate with Ethereum and with each other at faster finality. For Grvt, the important point is not that a trader’s balance will immediately talk to every institutional chain. It is that Grvt does not have to become a liquidity island. A user can trade in a private, high-throughput environment while the collateral and yield layer can eventually connect to Ethereum DeFi, tokenised funds and other ZKsync-native chains.

That matters because Prividium allows institutions to run their own private chains while still settling proofs back to Ethereum. Deutsche Bank’s DAMA 2 work with Memento uses ZKsync Prividium for tokenised fund management. Cari Network, a consortium of US regional banks, is using Prividium for tokenised deposits. ZKsync materials also reference the ADI Chain for sovereign-grade settlement infrastructure.

Atlas is what makes this ecosystem relevant to Grvt. If these chains can interoperate, then Grvt can become a venue where capital moves between trading, yield, and tokenised financial instruments instead of remaining trapped inside a single exchange. This philosophy is fairly different from what we see in traditional exchanges. The focus is not on having users trade more but building on infrastructure that can speak to a broader universe of assets.

In six months of interacting with the team, we observed a simple trend in where their focus sits. The obsession is not building yet another decentralised exchange that competes only on volume. Hypergambling does not work as a long-term business model because it leaves users worse off. Grvt’s focus has instead been on making new-age assets and primitives accessible to the average user without requiring the balances a conventional prime broker would expect. It has logical business reasoning too. The market for wealth-preserving products is not yet saturated, with over fifty players offering the same pitch of “get rich with crypto”. The cost of acquiring new users is drastically lower than a new exchange, where the primary focus is on speculation: larger TAM, lower CAC.

How does Grvt address these needs? For starters, consider that the average person does not sit online trading constantly. There is a market for that kind of product, but it is far smaller than the market for users who want a place to park their money, earn yield, trade the occasional narrative, and have access to their capital when they need it. Credit funds like the ones Blue Owl offers are not available to retail depositors if they have less than $100k. When they do need access to these funds for a personal emergency, the fund has the right to charge heavy withdrawal fees.

At the same time, these instruments are emerging on-chain at a rapid pace. Tokenised stocks are now available from Ondo. DTCC itself is considering native tokenisation of equities. As demand for AI-linked stocks from around the world increases, both awareness and demand for regional equities instruments will surge. Perpetuals and tokenised variations of equities are perfect for the world to buy into these waves. BlackRock, Apollo, and Franklin Templeton are in on the madness too. They have variations of money market funds and credit funds available on-chain today. According to RWA.xyz, there is roughly $1.4 billion of tokenised stocks, $14.6 billion of tokenised Treasuries and money market funds and $5.8 billion of credit represented on-chain. This excludes the trillion-dollar economy that stock perpetuals themselves have become in the past quarter.

A user from anywhere in the world can buy them for as little as $1 and receive yield on their capital. They also get to loop it on avenues like Morpho for additional yield.

We are witnessing a rapid expansion of the nature of assets that are available on-chain, and the unlocks it can offer users. But the marginal person is not looking to loop BlackRock’s securities fund on Morpho for a few extra points. They are not close enough to the ecosystem to track these shifts. The more important part here is that yield itself is commoditising. The moat is in distribution - or the sheer number of users a product can attract. And in how that capital base generating yield can speak to other product lines within the same product. We call this composability within our industry. And that is what Grvt has been focused on.

Why is that the case? A study of day-traders who traded consistently for more than 300 days in 2020 clarified that 97% of users lost money. Prediction markets are all the rage today, but 70% of the users on it lose money. The top 1% of users on Polymarket are responsible for 76.5% of the profits generated on it.

The average person on the web is not trying to escape the permanent underclass no matter how you cut the numbers. The TAM does not expand with better casinos.

These users instead park their money into the money market funds that are available on Grvt. The same funds can be used as equity to trade the occasional semiconductor stock or meme coin rally on the same platform. This blurring of worlds - between that of storage of wealth, access to tools to beat inflation and trade hot markets is what makes Grvt’s positioning unique. The way it works is rather novel - and it entirely relies on composability.

As of writing, the platform allows users to trade over 43 tokenised equity pairs - some of which include Korean equities that have been all the rage - alongside commodities like gold and crypto majors.

When a user parks money on Grvt for collateral, it is auto-converted into a yield-bearing instrument through the Earn on Equity product. Traditionally on an exchange, equity does not produce yield. The $100 you park to trade Bitcoin sits as $100 until you get liquidated or close a trade. Grvt allows users to earn yield on it. Earlier this year, they launched Aave as an integration so you could also verify where the yield itself came from.

For users that do not care to trade, Grvt also launched Grvt Invest. Instead of building these instruments in-house, they got Centrifuge and Plume involved. Centrifuge is responsible for onboarding the Janus Henderson Anemoy Treasury fund. Plume provides tokenised treasuries and BlackRock’s AAA-rated CLO ETF.

These instruments can also be used as collateral in vaults on exchanges. On Hyperliquid and Lighter, users park money in vaults to provide liquidity for trades. In exchange, they receive yield. Hyperliquid’s vault has $266M in TVL and has generated 12% APR. On Lighter, the APR is closer to 11%. The problem with such vaults is that they carry material risk. A vault takes on directional, market-making exposure depending on the market it is active in and can erode depositor capital during sharp moves.

The approach Grvt takes is interesting for two primary reasons.

1. They use tokenisation and composability as a wedge to offload the challenges of finding yield. Centrifuge is better suited to onboard Janus. Plume already has the BlackRock instrument on-chain. These are secure, stable instruments with regulatory cover. So instead of reinventing the wheel, they created demand for these instruments.
   

2. They also focus on giving users the ability to stack sources of yield. Yield can be competitive and high depending on the risk a user wants to expose themselves to. A user could park dollars and receive yield. Or buy into a credit fund to take some more risk. As of now, a user cannot provide capital into a vault, have margin, and earn yield all at the same time on most avenues. Grvt is working towards that vision by Q3.

Ultimately, the beauty of Grvt is that it can use blockchain rails to give access to niche financial products at a global scale through a permissionless mechanism with verifiability. If perpetuals were the embodiment of what makes blockchain rails an ideal capital movement technology, Grvt takes it one step further by using composability and tokenisation to build a product that goes beyond speculation. In an industry that focuses on having traders bet increasingly more, Grvt looks at how to help users retain more of their wealth and expand access to the instruments they can get hold of.

In many ways, this is symbolic of where crypto is. The tokenisation and protocol layer have reached maturity. What we are seeing now is a layer of fat apps that plug the best instruments and curate markets in applications. Such products will focus on distribution, curation, and retention as moats. In many ways, this rhymes with how wealth management in the traditional finance sector itself has evolved over the years.

When Everything is Tokenised

Crypto-native products have traditionally had a singular focus: to make users trade as often as they could, because the fee came from the transactions. If the broad arc of finance has a single message, it is that eventually you want to be the banking layer for the end consumer. The LTV of a user is substantially higher if the user parks their wealth within the product. Offering mechanisms to generate yield and access tokenised equities was not a possibility in 2022.

That reality has since changed. Grvt is the first of many products we will see transitioning to what I consider the “slow finance” phase of crypto. Or, in simpler terms, the coming of age.

In February, long before the yield-based products on Grvt were launched, I had a chance to speak to the founder, Hong Yea, about why any of this matters. The simple answer is that the market that seeks to store capital, access capital market products, and diversify asset bases while remaining self-custodied is substantially larger than the one for constant speculation. Cracks in equity markets are already visible today, with the KOSPI Index dropping as much as 12.6% on the 29th of July. We routinely see semiconductor stocks like SK Hynix and Samsung Electronics crash by close to double digits in a day. Having users hyper-gamble sounds like good business in the short run, but it does not lead to products that retain users.

More critically, there is another reason to build a retail user base for new money market instruments. Hong said conversion of users referred to the product jumped from 7% to 45% once the pitch shifted from trading to yield. Most people already understand the fixed interest rates their banks offer, so adding a few percentage points on top is an easy sell. Even then, Hong and the team may not earn most of their revenue from end users buying these instruments, but from issuers. Firms like Franklin Templeton and Blackstone have an incentive to pay Grvt for distribution of tokenised products once it reaches sufficient scale. Rather than relying on predatory tactics to push gambling products, Grvt is aiming to build the distribution layer that lets it work with the best asset issuers over time.

The distribution edge became more evident when I spoke to Ganesh Mahidhar from Further Ventures about their cheque into Grvt. His perspective is that Grvt has the opportunity to earn the entire stack for money. Users will be able to pay (via a debit card), access loans, earn yield, and access niche assets like computing derivatives on the same avenue. For tokenised asset issuers and mutual funds, the upside is access to a new distribution channel that, until 2024, did not exist due to the prohibitive costs of distribution, compliance, and servicing a customer at a traditional bank.

None of this is new conceptually. Crypto has pitched yield as a differentiator since at least 2018. For example, when crypto.com marketed higher-yield products in Singapore. What is different now is that the infrastructure has matured enough to support products like Grvt. Grvt is built on ZKsync and pairs that with its own risk engine for the markets it operates. ZKsync, in turn, can connect across multiple chains, where the on-chain “real-world asset” spectrum now ranges from stablecoins to tokenised real estate.

In that sense, the next big battleground for crypto applications is shifting away from base infrastructure (as it felt in 2022) and toward distribution, retention, and the business models that capture value over time.

The age of fat applications is here. This time, instead of being built centrally, with applications taking custody of user funds as FTX did, users have the option to hold their own assets. Even when they move capital into tokenised yield instruments, if Grvt goes down and the assets remain in their wallet, they can use a third-party interface to swap back into stablecoins. Grvt’s moat comes from its risk engine, brand, and distribution. Since the marginal user coming to Grvt is not a crypto-native trader looking for the 50th yield farm, retention should be structurally stronger.

It is rare for multiple primitives to mature at the same time in a way that enables a new product category. For Grvt to exist and scale, perpetuals had to be in demand, ZKsync had to mature, and tokenised instruments had to evolve simultaneously. The last comparable wave was when centralised firms were racing to offer yield, and FTX competed on a better perpetuals product. The primary difference this time is that we get to see these products built with radical transparency and self-custody. To me, that seems like a massive win.

Perhaps the pull of Gravity is not toward more speculation, but back toward financial sanity.

Playing with a fleet of agents,
Joel John

For the past year, DCo has been talking about how institutions are coming on-chain, but they’re coming on their own terms. They want privacy, compliance, and control, and they aren’t going to get those with products as they exist in their crypto-first form. Most of the institutional conversation has centred on infrastructure built by incumbents for incumbents, with Canton as the obvious example. We keep circling back to this question: can a crypto-native project, one that grew up in the open Ethereum world, actually serve institutions at scale?

ZKsync is one of the serious attempts to say yes, which is why it’s worth a close look.

This article walks through the stack and what Prividium actually does, why institutions might need their own chains and why they couldn’t have had them five years ago, how ZKsync compares to Canton in the race to become the institutional standard, and what the revenue model looks like across scenarios through 2028. If you were to bet on ZK, it has to be a low-probability, high-payoff bet, closer to an out-of-the-money option than a conviction hold.

TLDR;

• Institutions are coming on-chain on their own terms. ZKsync is one way to bet on a crypto-native project serving them. The market is not pricing in the optionality of its success.

• Teams have already built chains using ZKsync’s offering. These are live and some are thriving, which validates the stack’s technical capability. The enterprise narrative is a separate and still unproven claim. Prividium has institutional interest and early partners, but interest is not revenue. Whether it converts to paid production deployments is the open question.

• ZKsync offers a comprehensive suite for institutional blockchains. No competitor currently offers everything they need while anchoring to Ethereum. The bet is that the full stack under one roof matters more than the relationships incumbents already have.

Most of the value in crypto is on networks that won the distribution game, not necessarily the technology game. Arbitrum and Base control ~70% of L2 TVL. But two shifts can change the market dynamics and anoint new winners:

1. Institutions are coming on-chain for real this time.

2. ZK tech has finally caught up in cost and speed. ZKsync sits right at that intersection.

The retail story is almost settled with clear winners so far. The institutional side is chaotic, with a few winners emerging, but far from over. It is also messier, with different trade-offs. Institutions care about privacy and control more than cheap gas.

The numbers back the narrative. BlackRock’s BUIDL fund peaked at $2.9B in AUM in mid-2025. As of mid-2026, AUM has settled to roughly $2.0–2.5B across eight chains. It is tradeable via Uniswap through whitelisted qualified purchasers. This is BlackRock’s first direct integration with DeFi infrastructure. JPMorgan’s Kinexys settles more than $2B daily, with cumulative throughput past $1.4T. The Canton Network, backed by Goldman Sachs, DTCC, and JPMorgan, reports over $6T in tokenised assets and 15M+ monthly transactions.

Tokenised equities grew ~2,900% in a single year, from ~$30M to roughly $963M. Stablecoins now settle $9T+ monthly on a base of $300B+ in outstanding supply. And the DTCC, the entity that clears virtually all US securities, has received SEC authorisation to create blockchain-based digital twins of equities and Treasuries it already holds. The infrastructure question for institutions is no longer whether to come on-chain, but what stack to build on.

This is not the first time institutions have tried to use blockchain. Around 2017-2018, a wave of banks attempted to create their own private networks. Those efforts largely failed for three reasons: the technology was immature, the engineering complexity was beyond what banks could manage internally, and the chains they built were silos.

The next attempt was to directly use public networks like Ethereum. That failed because of compliance risks, regulatory exposure, customer privacy laws, and the simple fact that banks cannot have their balance sheets and trading activity visible on public rails. The institutions that tried the public-chain route faced regulatory consequences that made it untenable.

What Institutions Want and Who Can Give It to Them

Public-by-default blockchains clash with how finance works. Banks cannot expose trade strategies, client data, or counterparty relationships on a public ledger. It is about competition more than regulation. So, even if regulation allows it, banks wouldn’t want this part to be wholly public.

The four requirements institutions have articulated through ZKsync’s consultations with banks, corporations, and governments are:

1. Privacy (transaction-level confidentiality)

2. Control (governance over who participates and what they can do)

3. Risk management (predictable throughput, SLAs, and failover).

4. Interoperability (connection to global markets without bridges).

Institutions want a chain they control, privacy over balances and counterparties, and a way to reach public liquidity and settle on neutral ground. You can’t really have all of it at once. Almost everything on the market today gets you two corners and gives up the third.

Public chains and the assets issued on them, Ethereum, Base, BlackRock’s BUIDL, nail connectivity and settlement. They’re liquid and composable, but there’s no real privacy and you don’t control the chain. On the other side, you have bank-built chains like JPMorgan’s Kinexys and consortium networks like Canton and Corda. Those give you control and confidentiality, but they’re walled gardens with little reach into public liquidity, and their privacy leans on governance agreements rather than maths. Everyone has picked two and lived with the gap.

ZKsync’s whole pitch is that it doesn’t have to pick. Prividium lands in the middle of the triangle: a private chain the bank runs, confidentiality enforced by ZK proofs, and it still settles to Ethereum and can tap public liquidity when it needs to. That middle spot is the entire thesis. It’s also unproven. What it takes to actually get there is the rest of this piece.

Competition and the Standard War

Since the asymmetric part of the thesis depends on institutional adoption, the relevant competition is not just other public rollups. The most important comparison is the group of platforms competing to become the institutional standard for tokenised finance. The most relevant benchmark is likely Canton Network, not Arbitrum or Optimism. Arbitrum matters for retail L2 adoption. Canton matters for the institutional-standard race that underpins the non-consensus part of the ZK thesis.

Canton is a privacy-enabled L1 blockchain backed by Goldman Sachs, DTCC, Citadel Securities, BlackRock, Nasdaq, and S&P Global. Today, its incumbent ownership appears strong, and the ZKsync team is candid that this is their biggest competitive hurdle. Canton has deep, established relationships in the banking industry built over the years. ZKsync is the newer entrant that banks do not have existing relationships with. In institutional sales, that matters enormously because banks default to vendors they already know.

But Canton’s technology tells a different story. It is a fully permissioned chain that uses Daml, a proprietary smart contract language, thereby limiting the developer pool and creating vendor lock-in. A Flashbots deep dive on Canton flagged that membership is permissioned, participation is subject to institutional gating, and only approved financial institutions may operate nodes.

The same research flagged that a single entity controlling sequencers can achieve “God Mode” over sub-networks, enabling potential history rewrites or unauthorised participant admission.

The security models are philosophically different. Canton follows trust-by-policy. It has a known validator set, contractual governance, and consortium agreements. If the operator is hacked or hits a bug, there’s no cryptographic fallback. ZKsync, on the other hand, is trust-by-proof. The correctness of every transaction is verified mathematically.

Canton’s model is similar to how banks work today. It works well for bilateral settlement between known counterparties in a single jurisdiction with aligned incentives. What happens when entities that don’t know each other want to work together? Many institutions operate across borders, counterparties, and time horizons. A consortium with differing priorities and competing political pressures behaves correctly until it doesn’t. That’s where policy-based trust hits its ceiling.

ZKsync’s model has multiple layers of redundancy. The bank runs its own Prividium chain and can freeze or halt it if something goes wrong. Proofs are aggregated on the ZK Gateway before reaching Ethereum, catching failures before settlement. Even if the ZK circuits themselves have a bug, the system doesn’t collapse because each layer intervenes independently. No single operator can break the whole thing.

ZKsync’s other lever is EVM equivalence. Institutions building on ZKsync can tap into the existing Ethereum/EVM developer ecosystem and tooling. Canton’s Daml creates a closed ecosystem. And Ethereum itself is the most secure database in the world, with 33–34 million ETH staked (around 27% of supply) securing it. If companies want to transact with each other on a neutral, highly secure settlement layer, that would be Ethereum. ZKsync gives you native access to that.

The point is that institutions have to decide between Canton’s familiarity and ZKsync’s openness and Ethereum connectivity. Institutions tend to default to familiarity. When Canton just works, ZKsync will struggle to pull institutions towards an Ethereum-connected product.

Buying Its Way Into the Room

ZKsync is not ignoring Canton’s distribution advantage. The team has been aggressively countering it through several channels: a complete overhaul of marketing and branding since fall 2025; the Prividium Breakthrough Initiative, which brought Citi, Deutsche Bank, Mastercard, and other companies together to explore Prividium deployments; strategic partnerships with established players like Cari Network (five US regional banks with $600B+ in combined assets) and Bitco to leverage their existing banking connections; and direct engagement with government regulators in Washington DC.

The public debate between ZKsync and Canton has also intensified. Canton’s position is that ZK proofs “make everything invisible” while Canton “ensures the right people see the right data”; ZKsync’s counter is that an open ZK system gets harder to compromise as the verifier network grows, while a permissioned trust model stays as strong and vulnerable as its weakest operator. Nethermind formally verified the correctness of ZKsync’s on-chain ZK verifier using EasyCrypt in 2025, completing the first formal proof of this kind and directly addressing Canton’s concern about undetected bugs in ZK circuits.

ZKsync’s Comeback

ZKsync obviously didn’t start with this institutional approach. It launched as an Ethereum L2 chasing the same retail users as everyone else, and it lost that battle. Base and Arbitrum took the distribution, and Era is a rounding error next to them. ZKsync rebuilt itself into a full ZK stack (a chain, a chain factory, a prover, an interop layer, and managed services) and aimed it at institutions through Prividium, a private and compliant chain that still settles to Ethereum. The technology is now competitive. The Atlas upgrade pushed throughput past 15,000 TPS with roughly one-second finality, and the Airbender prover sits at the top of the public leaderboard on cost.

ZKsync is an enterprise infrastructure company that happens to have a public chain, ZKsync Era. You can think of the public chain as ZKsync dogfooding their own products to show what can be built. Its positioning is to be the institutional-grade ZK-rollup layer for Ethereum.

Ethereum’s execution market has bifurcated. Users and liquidity have concentrated on a small set of distribution-heavy rollups, while a second wave of validity-proof (ZK) systems is competing on performance, settlement certainty, and institutional fit. According to L2BEAT, different flavours of L2s secure ~$40B of total value. Arbitrum at $17B and Base at $11B lead the pack with ~70% of the total TVL. In ZK rollups specifically, ZKsync is still at the start. ZKsync Era secures ~$230M and is rated Stage 0, while Starknet secures ~$400M and is rated Stage 1. If one looks only at public chains, ZKsync Era has a long way to go to catch up and might never meaningfully close the gap, given how far ahead the leaders are.

But thankfully, it doesn’t have to. The story here is not about who wins over retail users. That space is arguably overcrowded and concentrated with Arbitrum and Base. At the same time, acquisition and retention costs are unsustainable and rarely convert mercenaries into durable users. The more interesting question is who will win where privacy, verifiability and institutional controls matter. ZKsync has gotten in on the ground floor of Ethereum enterprise adoption. But that doesn’t automatically mean durable distribution and recurring fees. We don’t know whether a technical edge is even what decides institutional deals in the first place.

ZKsync is implicitly competing in two markets: (1) crypto-native consumers and DeFi activity on a public rollup; and (2) the institutional market for tokenised assets, where compliance and privacy are the prerequisites.

• If ZKsync wins only as a public rollup, it competes head-on with distribution-heavy ecosystems (Base, Arbitrum, OP) where network effects already dominate. This is extremely tough, given that ZKsync has little market share.

• If ZKsync’s enterprise strategy works, it expands the TAM and changes the competitive set: permissioned environments, middleware, and tokenisation platforms. This is where the next leg of growth will come from, and ZKsync is well-positioned to capitalise on this shift.

What it Actually Does

ZKsync is a full‑stack ZK infrastructure platform. For an institution standing up a tokenisation product, it covers everything from spinning up a chain to running it day-to-day. It combines: (i) a chain factory for launching your chains; (ii) a network and interop layer that makes those chains interoperable; and (iii) a proving and execution layer that compresses proving cost/latency for better developer economics.

With this stack, ZKsync covers the spectrum from public consumer DeFi to app‑specific chains to regulated and private deployments (Prividium) while inheriting Ethereum’s verification.

Privacy & Enterprise stack

Prividium is the star product for this thesis. The asymmetric upside depends almost entirely on it. It’s a permissioned Validium (a Validium stores transaction data off-chain rather than on Ethereum) deployment built on ZK Stack, and it gives institutions three levels of customisation.

1. Privacy with auditability: Regular users don’t see each other’s balances, counterparties or transactions. But the chain operators (say, banks) can see everything so that they meet compliance requirements.

2. Enterprise-grade controls: This is a way to apply controls that banks already have at the blockchain level. Who can join the network, what can they do, KYC/AML checks, role-based permissions, SSO login through systems like Okta.

3. Ethereum settlement: ZK proofs from the chain get posted to Ethereum. This gives you external verifiability (anyone can confirm the chain is operating correctly) without exposing any of the private data. You get Ethereum’s security guarantees without making your transactions public.

In a way, Prividium lets a bank run a private blockchain that behaves like an internal ledger but settles to Ethereum. The bank controls who can transact, what they can see, and how compliance is enforced. All of this happens while still getting the security guarantees of Ethereum and the ability to connect to public liquidity when needed.

ZKsync Prividium is an Ethereum-anchored platform that provides users with private, permissioned, compliant blockchain environments with one-second finality. Optimistic rollups structurally cannot provide the same at this point. While Arbitrum and Optimism dominate current retail L2 TVL with optimistic rollups, ZK rollups have evolved to offer high throughput and structural advantages for enterprises. In short, it seems increasingly clear that ZK rollups will win over optimistic rollups (ORUs) in the long run:

1. Performance superiority: Airbender, ZKsync’s proving engine, ranked #1 on the Ethproofs leaderboard (January 2026) for cost per proof and #2 for proving speed. This is significant because proving used to be the bottleneck that made ZK rollups expensive relative to optimistic rollups. It no longer is. And ZK rollups still have a structural edge on finality: transactions settle in about a second, versus the 7-day challenge window that optimistic rollups require.

2. Institutional requirements: Enterprises require instant finality for settlement - they cannot wait 7 days for transaction finality that optimistic rollups require. As per the Prividium paper, many companies specifically chose ZK for this reason.

3. Privacy capabilities: ZK cryptography can hide transaction details by default. For institutions that offer confidential settlements for a multitude of use cases, that’s a hard requirement. Canton’s privacy model depends on governance agreements between participants (trust-by-policy). ZKsync’s model relies on mathematics (trust-by-proof). The difference shows up when participants operate across jurisdictions with different rules and different incentives. Policy-based privacy holds up as long as everyone cooperates. Cryptographic privacy holds up regardless.

Why this is hard, before we get to anything else

Selling core infrastructure to banks is one of the slowest and lowest-converting sales processes in technology. The cycles run for years through procurement, risk, compliance, and security review, and most pilots never make it to production. An MOU or a consortium initiative is just the start of a conversation, far from recurring revenue.

Banks also buy from people they trust and expect to still exist in a decade. ZKsync, or Matter Labs, is a startup that comes with entity and centralisation risks. Canton’s backers are its own customers. That relationship gap doesn’t usually get closed with better technology alone.

The one thing working in ZKsync’s favour is that switching and integration costs go both ways. They make the first deal hard to win, but sticky once it’s won. So the thesis really hangs on landing a handful of genuine production deployments, not on a wide funnel.

Interest Is Not Revenue

Prividium launched in May 2025. ZKsync’s conversations with institutions sit at different stages. Live and in production is one thing, signed is another, and exploring is the top of the funnel and by far the biggest bucket. The first named institutional partner is Deutsche Bank, which is co-building an on-chain fund management solution with Memento (a TradFi-to-crypto bridge company). In October 2025, ZKsync announced the Prividium Breakthrough Initiative, with Citi, Deutsche Bank, Mastercard, and 30+ global institutions participating. These are all exploratory engagements at the top of the funnel, not paid production deployments. In institutional infrastructure, most exploratory engagements don’t convert.

A US regional banks consortium called Cari Network selected Prividium to power a bank-governed tokenised deposit network. They have combined assets exceeding $600B. Unlike stablecoins issued by non-bank entities, Cari’s tokenised deposits remain FDIC-insured and on each bank’s balance sheet. This positions Prividium as the banking system’s answer to the $300B+ stablecoin market.

On the managed services side, ZKsync has 11 signed chains and 15 more in negotiation, with a pipeline of ~40 total prospects. Additional institutional partners include Securitize ($200M+ in tokenised assets on Era) and Tradable ($2B+ tokenised assets), exploring tokenised deposit networks on Prividium. The 2026 roadmap describes moving from foundational deployments serving tens of millions of users.

ZKsync’s technology is in production. The ZK Stack powers:

• GRVT: $85M TVL, $1B+ daily perp volume, processes 600K+ trades at under 2ms latency per trade. Mix of retail and institutional, with high-performance privacy-preserving execution.

• ADI chain: ZK chain licensed by the UAE Central Bank, powering the DDSC dirham-backed stablecoin. Live in production.

• Memento/Deutsche Bank: Prividium deployment under MAS Project Guardian (Project DAMA), co-building an on-chain fund management solution.

• Abstract: ~50K DAUs, consumer-facing chain (retail use case, validating stack versatility).

• Era itself: 700M+ total transactions, 291+ dApps, 162K daily transactions, $400M+ total value secured.

• 15 independent ZKsync chains are live in production on mainnet, with another 10 on testnet.

ZK Stack chains are already proving themselves in production. Perp DEX GRVT and consumer chain Abstract (L2 by Igloo Inc., Pudgy Penguins’ parent) have both used the stack to launch their chains. Abstract has ~50K DAUs, while GRVT shows the stack can support high-performance trading with $1B+ daily perp volume. These are not proofs of concept. They are live products handling real-world demands across multiple segments.

The most credible near-term commercial wedge is three specific use cases: 1) tokenised deposits, 2) treasury management, and 3) internal liquidity mobility. These solve an immediate, quantifiable pain point. A bank running a Prividium today can move liquidity between internal desks or across subsidiaries on private rails that settle to Ethereum, without waiting for the rest of the financial system to coordinate around a shared standard.

What's It Worth If It Works

The market currently values ZKsync as a failed retail general-purpose L2 and prices in little or no probability that it will become institutional infrastructure. The model below sizes the payoff.

ZKsync is one of the closest candidates to be Ethereum’s answer for enterprise chains, and it’s been moving in that direction. On that basis, we examine the various revenue streams and estimate how much ZKsync could generate through the end of 2028.

ZKsync’s product stack has three broad user categories: 1) developers, 2) crypto-native chain operators, and 3) institutions. Each enters the ecosystem at a different layer and pays in a different way.

Developers build on Era or use the SDKs to generate transaction fees. Chain operators deploy their own chains using ZK Stack and pay for Managed Services subscriptions to handle sequencing, proving, and data infrastructure. Institutions come in through Prividium and pay enterprise licensing fees for private, compliant blockchain environments coupled with managed services.

All the potential users or customers have different entry points into the stack. But all of them have the freedom to move across the stack once they are in. A chain operator that starts with a basic ZK Stack deployment might later upgrade to Managed Services, and an institution evaluating Prividium will likely need managed infrastructure bundled in. Each layer creates a fee rail that compounds as adoption deepens.

Where the Money Comes From

Each of the six possible revenue lines above is about a specific part of the ZKsync product stack. Here is what drives each one and why we think it can scale.

1. Enterprise Prividium will decide ZKsync’s fate, and it’s also the least proven. Institutions need private blockchain environments that are also compliant, and Prividium is one of the few products that deliver privacy, access controls, and Ethereum settlement in a single package. Institutions can either deploy through Prividium and run it themselves or hand it to ZKsync via managed services. Six Prividium deployments are targeted to be live by the end of 2026, though that’s a target, not a commitment. The early use cases here are tokenised deposits, treasury management, and bank-to-bank settlement. If those work, they open the door to repo, FX, and securities settlement. Given the scope, this is the line that makes or breaks the thesis.

2. Managed Services is the steady-state subscription business. As mentioned earlier, if you would rather not run a chain, ZKsync does that for you. Teams that want to run a ZK Stack chain without managing sequencing, proving, RPC, indexers, and data services themselves can pay an annual subscription. The subscription will be a cheaper option than running DevOps by yourself. This shifts ZKsync from cyclical on-chain fee revenue to more predictable recurring income, with room to expand through premium add-ons like private DA, dedicated regions, and priority support. There are already 11 signed chains and 15 more in negotiation.

3. Era Chain Fees are the most straightforward line: user transaction fees on ZKsync’s public L2. This is live today, but it is not the primary growth driver. Era serves more as a proving ground for the technology than as a revenue engine in its own right. That said, with 700M+ total transactions and 291+ dApps, the stack demonstrates it works at scale.

4. Interop / Gateway Fees become relevant with scale. ZKsync Connect and Gateway went live in June 2026. The idea is that every cross-chain transaction routed through the Elastic Network pays a small protocol fee. As the volume of cross-chain messages grows, ZKsync can charge a small fee per message.

5. Elastic Chain Network Fee is speculative based on Arbitrum and Optimism stacks charging 10% on the net revenue and 2.5% on gross revenue, respectively. Although ZKsync doesn’t currently enforce this, as the Elastic Network grows, it remains a credible future lever.

6. Proving Services is also speculative. Not every chain that wants ZK proofs will build its own prover. Airbender currently sits at #1 for cost per proof on the Ethproofs leaderboard and at the top tier for proving speed. If proving-as-a-service becomes an industry norm, ZKsync is well-positioned.

The table below maps each revenue line to market comparables and an estimated revenue range.

ZKsync could generate about $4-7 million in revenue in 2026. The largest driver for this would be the enterprise Prividium deployments. If six Prividium deployments are live by the end of 2026, with an average revenue of ~$400k-$600k, they will generate $2.4 million to $3.6 million. Managed services are the second pillar. 15 of these at the end of 2026 will amount to ~$1.5 million to $3 million. The Era chain and interop gateway are unlikely to be heavy contributors this year. The remaining two streams will only start contributing once other lines solidify.

We would normally take the 2026 estimate, apply a multiple, and assess potential valuations. But we can’t do that because ZKsync’s institutional push began around the second half of 2025. Given how time-consuming the sales process is for banks and financial institutions, it will take at least three years to validate the ‘enterprisification’ thesis. This is why we built the model out to 2028. The table below shows assumptions used for 2028 revenue contributions across all revenue streams.

We bring those revenues back with a 20% discount rate. So this is what revenues under different scenarios look like.

So where does this leave us? Step back from ZKsync for a moment. The trend that keeps winning in crypto is tokenisation, not token issuance. We wrote about this recently. Investors have stopped paying up for low-revenue governance tokens, and value has moved to assets with real revenue, real moats, or plain equity. Stablecoins were the first wave, and on-chain stocks are the second. The logical progression is that blockchains become the place where the world’s assets are issued, traded, and settled. And these assets will mostly be issued by large institutions.

What’s clear is that institutions will use blockchains. But we don’t know how or which blockchains. Do they build their own chains, adopt crypto-native infrastructure like ZKsync, or stick with the familiar vendors already in the building? And how big a shift is this really?

Moving settlement and record-keeping from fully centralised systems to a more decentralised version could be a generational re-architecture of financial infrastructure, comparable to the move from paper to digital. Or it could end up as a ‘not-a-big-deal’ upgrade that mostly preserves the existing power structure. Then there’s the question that decides whether any of it holds together: once a dozen institutions are each running their own environment, how do those systems talk to each other? We don’t have answers yet.

Our own view is that this is a bigger change. Can crypto-native projects be useful to this tokenisation trend? The time and environment feel right for it. The next phase of crypto is about bringing the world’s best assets on-chain, and that doesn’t happen without institutions. Whoever builds the rails they settle on and connects those rails to each other captures the most value of this cycle. ZKsync is one of the more credible crypto-native attempts to serve as that layer for Ethereum, which is why it’s worth watching. The incumbents have the relationships; ZKsync has the architecture and the bet on open networks. Which of those matters more is what the next two years will decide.

Exploring how crypto expands finance TAM,
Saurabh Deshpande

Institutions are here. They are buying our assets, they are tokenising their assets. But they are not fully on-chain in a true sense. Institutions still use their assets off-chain and use blockchains for registry and cheaper, continuous settlement. We think the lack of on-chain privacy when using the asset has been a major deterrent to players moving active capital on-chain.

Starknet is changing that by ensuring that you can use your assets while maintaining privacy. This essay, written in collaboration with Starknet, is about why we think more capital will be on-chain and put to use there once the missing privacy puzzle is solved.

TL; DR:

• Transparency is the trap. Transparency became crypto’s institutional bottleneck. A market needs the trade, not the trader. Public chains exposed everything to everybody.

• That locks institutions out. Funds, treasuries and custodians may use public rails in narrow cases, but they cannot operate at full depth where every position, payment and counterparty relationship becomes public strategy.

• Every previous solution broke a critical component of the market. Most solutions could not get privacy, liquidity, and compliance right. Zcash and Monero created private islands, Solana-style confidential transfers provided partial privacy, and Tornado Cash invited regulators.

• STRK20 is betting that it is better to make existing assets private. USDC, tokenised treasuries, BTC wrappers and stablecoins can become private without becoming new assets with separate liquidity.

In the late 18th century, Jeremy Bentham designed a perfect prison called the Panopticon. The design stacks prison cells around a single watchtower. Every cell is lit from behind so that guards in the watchtower can see into each cell at once. While no inmate could ever tell whether anyone was up in the tower looking back. They never knew when they were being watched, so they behaved as if they always were. Two centuries later, the philosopher Michel Foucault seized on the design as the template for modern power. He said that visibility is a trap. You do not need chains if everyone can be seen. Watching alone does the work.

Crypto built a Panopticon without guards. It just handed every prisoner a pair of binoculars and pointed them at each other.

Don’t trust; verify. That was our industry’s go-to line. For a long time, a public ledger’s transparency was the selling point. More than anything, we hoped that the world would somehow be okay with putting its finances under everybody’s scrutiny.

The hope remained for quite some time. After a decade of opaque intermediaries, blown-up exchanges and rehypothecated claims, a ledger anyone could audit felt like the fix. You could check balances, verify supply and confirm a trade settled without trusting an operator. Transparency was not one feature among many. It was the whole pitch.

The problem is that if crypto wants to attract traditional finance players, it must contradict that original pitch.

The ambition has moved beyond retail speculation to all of finance. Tokenised treasuries, stablecoin settlement, on-chain credit, bitcoin collateral, corporate treasury flows, institutional DeFi. Crypto wants the serious money, and serious money does not want to be watched. They are not doing anything wrong. It is just that they might be hurt by moving money in public.

So, crypto is in a strange place. To win the institutions it has been chasing for years, it must rebuild the one thing it spent a decade defining itself against: privacy. The kind of privacy that every serious financial system already has. The one with regulatory oversight.

One might think this is a retreat from transparency. To an extent, it is. But it is also the realisation that transparency and confidentiality were never the same thing. Early Bitcoin proponents were also advocates of privacy. Ironically, Bitcoin is one of the more transparent systems out there. The thing is, markets need to know that a trade happened so the price can be updated. They have never needed to know who made the trade. Public blockchains fused those two facts into a single data layer, and that fusion stands between crypto and the capital it wants.

Starknet has spent the past few years resolving this tension between transparency, privacy, and compliance. It was designed around provable execution, in which a transaction can be shown to be valid without exposing its contents. This design allows Starknet to embed privacy in the core of the stack rather than add it as a feature. STRK20 is the expression of that idea. It gives any asset already living on Starknet a private mode without reissuing it. And does it while still trading against the same public liquidity everyone else uses and keeping a lawful path open for oversight.

This article is about how crypto can resolve the tension between privacy and publicly verifiable settlement at low cost. The aim should be to rebuild the invisible parts of the financial market structure without abandoning public settlement. This ensures that financial market participants do not have to jump through hoops when they use blockchain rails.

The Cost of Being Seen

Public blockchains are like a transparent fishbowl where everything is visible to everyone, all the time.

Traditionally, finance has separated what the market needs to see to function from what participants are allowed to protect. The market needs prices, liquidity, finality and proof that a trade was settled. There is no need for a perpetual live feed of who is doing what.

Take Warren Buffett as an example. Berkshire Hathaway has often sought confidential treatment from the SEC to build positions. This way, the market does not get the opportunity to front-run it. Phillips 66 in 2015, Chevron and Verizon in 2020, and most recently Chubb, where it quietly accumulated a roughly $6.7 billion stake through 2023 and 2024 before disclosing it in May 2024. The SEC grants this to large investors from time to time so that they can accumulate shares at a reasonable cost basis. The mechanism exists because information moves prices.

Dark pools exist for the same reason. The SEC describes them as alternative trading systems that operate without displaying specific order information prior to trades. They exist so that a large order does not move the market before it fills. They are not a fringe venue. Dark pools run roughly 15 to 18% of US equity volume on their own, and once you count every off-exchange venue, including bank-run pools like Goldman Sachs’ Sigma X and JPMorgan’s JPMX, more than half of all US equity trading now happens away from public exchanges. FINRA publishes the off-exchange data because it is now an integral part of the market structure.

A public chain has none of this machinery. There is no confidential treatment you can apply for, no dark pool to route your orders through, and no delay before disclosure. When a wallet starts buying in size, it often alerts the market before the order gets filled.

Neither Berkshire’s confidential filing nor the dark pool is the natural state of the market. Both are additional infrastructure built on top of a system whose default is full exposure. They are also gated and rationed. When you place a large order on a public exchange, the book shows your size to everyone, though not your name. Confidentiality in traditional markets is engineered, and access to it is sold to the people who need it.

Crypto is yet to build engineering infrastructure that hides things the way traditional markets do. It makes the exposure worse because when one of your addresses is linked to your identity, it becomes difficult to maintain pseudonymity. Of course, sophisticated users know well to keep changing their addresses often. But they must be perennially vigilant. On a DEX, your trade waits in a public mempool, which means you announce what you are about to do and then wait a few seconds for someone to act on it. And bots do. They buy in front of you and sell into the price your own trade pushes up.

Every serious market is eventually forced to build a confidentiality layer on top of its public one. Because without it, capital will not move in size. Traditional finance built that layer as separate, gated venues. Crypto must build it too. The only question is whether it can be done without walling the private flow off into a room of its own.

The lack of privacy in crypto is not a bug. The openness was deliberate and the whole point of the design. Reading the ledger yourself, trusting no one, was the pitch. And it only works if everything is visible to everyone.

Legibility has never been neutral. In Seeing Like a State, the anthropologist James Scott showed how governments learned to control a society by first making it legible. You cannot control what you cannot read, right? So, they flattened messy local idiosyncrasies into easy-enough-to-read from desk records. Cadastral maps in place of tangled local land claims. Standardised weights. Permanent surnames so people could be taxed and conscripted. Scott argued that whoever can access and read the map ends up controlling the territory.

When you think of it, crypto built the most legible market map in all of finance history. But unlike any state, it handed the map to everyone. Instead of one entity reading the ledger, the whole network does. Just that in this case, the most sophisticated reader wins.

Curve was one of the largest DeFi protocols. In June 2024, Michael Egorov, the founder of Curve, had borrowed close to $96 million in stablecoins against about $141 million of his own CRV tokens. The activity spread across several lending markets, such as Inverse, UwU Lend, Fraxlend, and Curve’s own Llamalend. Since all of it lived on-chain, anyone could read how much he owed, what backed it, and the exact CRV price at which each loan would be liquidated. George Soros’ reflexivity 101 tells us that when a major liquidation point is known, the market usually hunts for it. CRV started sliding as people began to front-run the liquidation. The positions unwound in a cascade that wiped them out, leaving around $10 million in bad debt. A public liquidation price is not essential information to the market’s functioning. It is often a target.

In a small market like crypto, this level of transparency was okay. But if you are a mutual fund where millions of people put money via SIPs, you do not want other funds to know what and how much you are buying. No one with size wants to be in a transparent fishbowl.

Why Privacy Kept Failing

Privacy is not a new idea in crypto. The industry has tried to achieve it many times. Each attempt tried to solve the predecessor’s problems but broke differently. We will look at three prominent approaches here.

Isolated liquidity

The first approach was to make privacy native to the base layer. And this was the correct instinct. Zcash and Monero built privacy into the base layer, so a transaction was private by default or close to it. Although the cryptography was good, everything around it was the problem.

The lack of liquidity was the first major issue. When an asset is private on its own island, it must build the entire financial stack there. Without exchanges, lending markets, fiat ramps, and custodians, it means nothing. But why will anyone build these things without liquidity? Without the financial plumbing, the private asset has nowhere to go. So, achieving scale was difficult.

The second problem was that users had to take action to make their assets private. Others could watch assets entering and exiting the privacy pools. It is like entering the event horizon. People can see you until you get in. Capital markets cannot work with strong privacy and weak access.

Zcash made shielding optional. Users had little reason to park their assets in the privacy pool. So, the crowd never formed. When the pool is relatively empty, it becomes easier to ascribe assets to their respective owner. One study reconstructed roughly a third of all coins ever sent to shielded addresses, just by matching amounts, timing, and the steady traffic between shielded and transparent accounts. There was simply nobody to hide among.

Another problem with this design is that when you hide everything, you hide what you need to verify as well. In 2018, an Electric Coin Company cryptographer found a flaw in the zk-SNARK construction behind Zcash’s shielded pool. An error in the setup allowed a cheating prover to forge valid-looking proofs, but it also allowed an attacker to mint unlimited counterfeit shielded ZEC. The invisibility made the bug more frightening. Because the shielded supply was private, the counterfeiting would have left no trace. You cannot audit what you cannot see. Zcash patched it quietly in late 2018 and disclosed it the following year, and there is no evidence anyone ever exploited it (Electric Coin Company).

Zcash’s trusted setup had a bug. Think of a trusted setup as a group of people gathering around one secret ingredient. They mix their individual secrets together to create a master key. If any participant had kept their share of it, they could have forged proofs forever. The safety of the whole system rested on the bet that at least one of them had destroyed their secret.

A shielded Zcash transaction really does conceal its sender, receiver and amount. The catch is that hiding the contents of a transaction does not hide the transaction. An observer still sees that a shielded transfer happened, when it happened, and the amounts going into the shielded pool and coming back out. Turning those visible facts into an identity only requires that there be few enough candidates. If a thousand people shield similar amounts at similar times, yours is lost in the pile. If three people do, you are one of three. Cryptography cannot fix this. You have exactly as much anonymity as there are people you could be confused with.

Partial privacy

The second approach kept the public chain and hid only part of the activity on it. Solana’s Token-2022 confidential transfer extension is the clearest example. It encrypts an account’s balance and the amount of a transfer while the token remains on Solana, shipped as part of Solana’s Confidential Balances suite in 2025, with a built-in auditor key so a regulator can still decrypt it when required.

A year on, it sits almost entirely unused. PayPal’s PYUSD can support confidential transactions. However, major wallets like Phantom and Solflare still do not support it natively; the proofs are generated off-chain, and an account must be set up with an encryption key before it can receive a confidential transfer at all. It exists as a feature without much ecosystem support.

Confidential transactions hide the amount but leave the remaining transfer parameters, such as the accounts, counterparties, and timing, public. And it does not compose with other primitives like swaps or lending. A confidential transfer is only a transfer. The moment you want the money to do something, swap it, lend it, post it as collateral, the protocol on the other side needs a number it can act on. So, the figure you encrypted is decrypted the instant you use it, which leaves you with privacy that holds only while the asset is passively held. Almost nobody wants privacy that they must pay for with their own money.

Mixing

The third approach was to break the trail by combining everyone’s assets. Tornado Cash lets users deposit on Ethereum and withdraw to a fresh address. Because users transferred from their addresses to the pool and then to new addresses, the on-chain link between their addresses was severed. Instead of asking people to leave Ethereum, it simply interrupted traceability by adding a black box in between.

The US Treasury sanctioned Tornado Cash in 2022, alleging that over $7 billion had been laundered through it. It also included funds tied to North Korea’s nefarious Lazarus Group. A federal appeals court ruled the sanction unlawful in late 2024, noting that immutable smart contracts are not property the government can sanction, and Treasury formally delisted the protocol from sanctions in March 2025. Despite that, one of its founders was convicted in August 2025 of running an unlicensed money-transmitting business, while the jury hung on the heavier money-laundering and sanctions charges.

Although you can interpret the whole thing as a win for privacy, institutions would not care for such a victory. The sanction was reversed, but only after years of litigation, a criminal trial, and a developer facing charges that carried a sentence of up to 20 years in prison. No financial institution wants to deploy capital and wait for the law to eventually catch up.

There was one more problem with Tornado Cash. You could “anonymise” your ETH in pools with different denominations of 0.1 ETH, 1ETH, and so on. The higher you went, the more difficult it was to hide because there was not enough of a crowd.

From all these attempts, it is clear that privacy needs to satisfy the following criteria.

1. It cannot be on an island where the whole finance stack must be rebuilt. Expecting users to bypass the existing infrastructure is a terrible UX. So, existing liquidity cannot be broken.

2. You must ensure that the solution complies with the existing regulations. Without regulation, you cannot achieve critical mass and hoping that regulation comes around is a terrible strategy.

Each one solved a piece of the problem and broke on another. Institutions want privacy, market access, and a compliance solution all at once. No system offered all three.

That is the gap Starknet is aiming at.

Proving it, Instead of Showing it

Eli Ben-Sasson, co-founder of Starknet, also co-invented Zerocash, the protocol underneath Zcash. He had seen first-hand how strong cryptography could still leave money stranded. Privacy cannot be a layer you bolt onto a transparent system. It must be a property of the system.

To see why that matters, let’s start with how an ordinary chain checks that a block is real. On Ethereum, when you send 100 USDC, every validator checks the transaction by re-executing it locally. It is like a room of accountants balancing assets and liabilities of a balance sheet from the income statement. If validators must check everything, they must have all the information. Re-execution is the reason the entire state must be public. You cannot recompute what you cannot see, so transparency was never really a choice those chains made. A consequence of this is also that privacy cannot be added afterwards, because hiding the numbers would blind the very accountants the system relies on.

Starknet does not ask every node to redo the computation. One party runs it and produces a STARK proof, and everyone else just checks the proof. Re-running a computation to check it requires its inputs. But verifying a proof has no such requirement. The validator confirms the result is correct without knowing any values. So, correctness no longer depends on disclosure.

Starknet distinguishes between two things: whether a transaction is correct and whether its contents are visible. As a result, privacy can be at the base layer.

In May 2026, a researcher auditing Zcash’s newest shielded pool (called Orchard) found a soundness bug in its zero-knowledge circuit. It had been live since Orchard launched in 2022. It could have been used to counterfeit ZEC inside the pool. The Zcash Foundation disabled Orchard with an emergency soft fork and patched it two days later. There is no sign that suggests anybody is exploiting the bug.

However, Zcash cannot prove that. The balances inside the Orchard pool are hidden, so there is no way to open the books and check that no coin was forged. The bug could forge coins in the pool but removing them would be difficult. Zcash runs a turnstile on every pool, a rule that tracks what goes in and never lets more come out than went in. So, however much was forged inside Orchard, the pool could not release more ZEC than people had deposited, and no counterfeit value could reach circulation. That check holds only because the pool’s deposits and withdrawals are public. No one can be certain that the balances inside Orchard still add up.

A hidden supply makes a bug like this invisible and unprovable after the fact. It was not the old kind of failure. Orchard runs on a proof system with no trusted setup, so there was no ceremony and no toxic waste to blame. The bug was a mistake in the circuit, which is an integral part of the proof system. Removing the trusted setup did not remove the danger.

This is why privacy must be close to execution. When you add privacy to public systems, they leak information at junctions. Say you have a private wallet that is compatible with the public chain. There would need to be a bridge mechanism to move your assets from regular addresses to private ones. The bridge is likely to expose asset movement. Privacy is tougher to break when it is native to the chain, i.e., inside the execution.

The same machine for privacy

Programs on Starknet are written in Cairo, a language built so that everything on it produces a mathematical trace that can be turned into a STARK proof. The same proving system that verifies a whole block also proves a single private transaction. Both are just STARK proofs that Cairo code runs correctly, so the chain handles them with a single set of machinery. Unlike in earlier cases where privacy was an add-on, the privacy engine is treated like the rest of the system components.

Starknet uses STARKs rather than the SNARKs Zcash used for two reasons. The first is that STARKs do not need a trusted setup. A SNARK begins with a one-time ceremony as mentioned before. Although Zcash ran elaborate ceremonies to lower the risk, and one participant destroyed their hardware with a blowtorch, the assumption that there is at least one honest participant must hold.

STARKs do not need this assumption. No ceremony, and no master secret for anyone to hold. But the setup secret is only one of two ways a shielded system can be forged. The other is a bug in the proof circuit itself, a flaw that lets an invalid transaction pass as valid. This has nothing to do with the trusted setup.

The second reason is the longer horizon. SNARKs rely on elliptic-curve cryptography, which a sufficiently powerful quantum computer could eventually break. STARKs use only hash functions, which are quantum-resistant to attacks known today. A system designed to keep financial records private for decades must address quantum threats. Besides, governments and banks are already moving to post-quantum standards.

With the architecture and cryptography now in place, one piece was still missing. For privacy to be genuinely useful, it needs to work on all devices. A good design does not make users spend a lot of money. So, users should be able to build proofs on their devices without handing them over to servers.

A single STARK proof is tens of thousands of numbers. Checking one such proof required running it through a smart contract. But the proof was too bulky for a single transaction, so verification was slow and costly. The answer was to somehow make users do less work. So, the April 2026 Shinobi upgrade moved verification out of a smart contract and into the network’s own consensus. This transferred verification work from the user/developer to the chain validators. So, the chain now checks a privacy proof as routinely as it checks a block. A private transaction that could not be verified at scale before now settles in seconds. Everything else in this piece depends on that change.

What STRK20 does

STRK20 gives an asset that already exists on Starknet a private mode. You do not have to mint a new asset. A stablecoin, a tokenised treasury, wrapped bitcoin, or any ERC-20 token can enter a shared privacy pool and gain confidentiality. Nothing is reissued, forked or migrated. The liquidity and the protocols already on Starknet can become privacy-capable from where they are.

The pool behaves like a drawer of sealed envelopes. When you deposit funds, they become a note. It is a private record of how much you hold and a secret that only you know. That note is sealed inside an envelope with what cryptographers call a commitment, and the envelope is placed in a drawer shared by everyone in the pool. Anyone can see the drawer is full of envelopes. No one can see inside any of them.

When you want to spend your assets, you do two things, both on your own device. You prove that you hold the secret to one of the envelopes in the drawer, without revealing it. And you publish a nullifier, a short marker built from that note’s secret. One note can produce only one nullifier, so publishing it marks the note as spent. The nullifier cannot be traced back to any envelope, so it retires your note without showing which one it was.

This prevents anyone from spending the same note twice, even when no balances are displayed. There is no balance to check against, so the chain instead keeps a public list of every nullifier ever published. Spend a note once, and its nullifier joins the list. Try spending it again, and the same nullifier is already there, so the network rejects it. A bug in this exact machinery is what cost Zcash its soundness, more than once.

The full cycle for a user’s asset is deposit, hold, transfer, trade, withdraw. A transfer inside the pool hides the sender, the receiver, the amount, and even which asset moved. To everyone else, it is just another entry in the pool. The pool is built to hold many assets at once, so one person can keep USDC, wrapped BTC and tokenised treasuries together, with no one able to read the portfolio. And because account abstraction is native to Starknet, the design carries over multisig, smart accounts and social recovery from the start, so an institution keeps the security model it already runs on rather than trading it away for privacy. The protocol behind it is set out in its technical paper.

A transfer only concerns two parties. A swap must touch a shared market. How does a market that knows nothing about a trade reprice against it? Every earlier system broke right here. STRK20’s answer is to control what each party in a swap can see.

When you swap privately on avnu or Ekubo, the trade performs as a single atomic action against the same liquidity as public swaps. The pool observes the size and direction of the trade because it needs those parameters to adjust its price and reserves. It does not care about who made it, what else they hold, or how this swap connects to their previous one. Your order never waits in a public mempool. So, there’s no question of frontrunning it. Unlike other attempts that split liquidity into several private buckets, STRK20 maintains a single shared market while hiding the participant. This way, the market gets the price it needs but never gets the person or address behind it.

A key for the regulator

None of this matters if regulators come for it the way they came for Tornado Cash. Crypto-natives tend to read compliance as surrender. The instinct is principled and naive about how regulated finance works. A custodian, a bank, a fund or a payment firm cannot opt out of oversight. It needs confidentiality from the public and its competitors. It cannot get immunity from the law, and it is not asking for it. Your bank does not publish your transactions, yet a court can compel disclosure, and an auditor can inspect the books.

STRK20’s answer is a viewing key that each user holds. The key can decrypt only its own history. Think of a safe-deposit box that only you can open. When you are renting it, you must lodge a sealed spare key with a custodian. The custodian is a threshold-controlled group of independent auditors. On a lawful request, they can use the spare key to reconstruct a user’s trail. They cannot open the entire pool. At the entrance and exit, wallets are screened for sanctions lists, so tainted funds are turned away at the door.

Entry to the pool is gated by a proof that the viewing key was escrowed correctly, and there is no way to skip it. This means that compliance is built into the act of using the system. You cannot use the pool without being compliant. Financial institutions do not have to worry about auditing compliance promises of every downstream on-chain product. They have one defensible answer they can give to a regulator, and a rule enforced by the protocol is far easier to evaluate than a stack of app-level discretion. It lowers the cost of diligence.

However, the design is a compromise between ethos and what works in the real world. A spare key held by anyone other than the user reintroduces a trusted middleman. This is one of the same centralisation vectors crypto set out to avoid, and a purist who calls it a backdoor is not entirely wrong. The system’s integrity does, to some extent, rely on how the authorised party handles the key. All the banks and financial institutions use some version of this compromise. It is narrower than trustless privacy. What STRK20 offers is market-independent privacy, with a lawful, scoped, and enforced oversight route. It is what an institution can use.

Market for Privacy

Every privacy system before this one has offered some of what an institution wants. None has offered it all at once. The wish list is short with stubborn trade-offs. It should –

1. Be private from the market,

2. Be able to trade against shared public liquidity,

3. Have a lawful path for oversight and

4. All of it should work with assets people already hold, rather than requiring people to mint new coins.

Zcash and Monero deliver privacy and little else. The asset is private, but it is also its own island, with its own thin liquidity and no oversight route an institution could lean on. Strong on one axis, absent on the rest.

Solana’s confidential transfers keep you on a public chain with real liquidity. But they hide the balance and leave the act of trading in the open. The amount is masked, while the behaviour around it remains legible, and there is no built-in audit trail. Good access, partial privacy.

Aztec is the most serious privacy engineering in the space. It runs fully private smart contracts on its own Ethereum L2 with client-side proving, and it has been live since late 2025. Privacy comes with a caveat. It lives inside the Aztec environment, which means the liquidity does too. A private Aztec user can only trade against the liquidity already existing in Aztec’s ecosystem, thereby prohibiting access to deep public liquidity outside it.

Canton has the institutional support. It is a permissioned network built for regulated players, with privacy and compliance standards similar to those that banks already maintain. Its strength also becomes its ceiling. It is a walled garden of approved participants, and private capital cannot enter or exit it at will.

Railgun is the closest cousin to what STRK20 is attempting. It lets existing assets on public chains move into a shielded pool, use DeFi privately, and even prove funds are clean through a tool it calls Private Proofs of Innocence. It works today. What it lacks is scale and speed. It runs on zk-SNARKs with the trusted setup baggage that implies, routes trades through relayers, and stays small enough that the anonymity on offer is thinner than the pitch.

Zcash put privacy on its own chain, so its liquidity got stranded. Solana added privacy on top of its transparent base layer, so it leaks the moment you trade. Aztec built a private world of its own, so its market is only what has moved there. Canton stayed permissioned, so it is closed by design. STRK20 puts privacy at the foundation. For the first time, you do not have to move elsewhere if you want privacy. We must wait and see whether these advantages help drive genuine adoption.

Can privacy bring new demand?

Most privacy projects carry a hidden tax. If the privacy lives in a new place, every application must be rebuilt there. STRK20 tries to avoid tax by making DeFi on Starknet privacy-capable. An asset can move into private mode and still reach the apps that exist today, so the existing market becomes more useful rather than being abandoned for a new one.

Developers do not want to add privacy for its own sake. It is no good if it does not bring additional users or more volume to their app. Can privacy drive net-new demand? The order flow that matters most is the kind that cannot come on-chain at all today. A fund that will not reveal its positions, a company that will not publish who it pays, a custodian that cannot let one client see another’s book, none of it can touch a transparent chain. A DEX or a lending market that turns privacy-capable can open the door to the capital that has been standing outside.

What Starknet is Trying to Solve

So far, this has been about the design. But who are the likely candidates to use this? Of course, crypto native users who have already used on-chain finance. But it expands to the market to the institutions that have circled on-chain finance for years and never stepped in, because the rails would have put their every position and payment on display. For most of them, the asset can already be on-chain. With baked-in privacy, just what they are used to in the traditional world, they can now start using on-chain assets.

Tokenised and exposed

The value of tokenised real-world assets on-chain, everything except stablecoins, grew from around $6 billion in early 2025 to more than $30 billion by mid-2026. BlackRock’s tokenised treasury fund, BUIDL, now serves as collateral in DeFi.

$30 billion is a rounding error compared to the tens of trillions in the global fixed-income market. Of course, many factors, such as regulation, distribution, and the absence of an ecosystem, explain why the entire thing is not on-chain. However, an asset manager will not run its book on rails without privacy.

So, the tokens get issued on-chain. However, the things you would do with those tokens tend to stay off- chain. It is good only for those who did not have access to these assets at all. Off-chain rails are still better for those who already use these assets. This can only change when a tokenised asset holder can trade without publishing the strategy behind it.

Where stablecoins stall

There’s now no doubt that stablecoins already work. They move value across borders, settle in seconds and run outside banking hours. Of the ~$39 trillion in annual transfer volume, genuine payments are about 1%, or $390 billion. The B2B payments share has grown more than 700% in a year to around $226 billion.

The 700% is real, but the volume is insignificant compared to the $89 trillion in global B2B payments. Stablecoins have managed to partially capture lower-hanging fruit like cross-border transfers between crypto firms, settlement for exchanges, and payments small enough that no rival is watching. The valuable flows stay away because they come with a cost to be seen. Companies like Kellogg or Mars are not going to want the world to know who their suppliers are by using on-chain payments for convenience.

Speed and cost, the original USPs of stablecoins, cannot come at the cost of privacy. With confidentiality, they become rails for serious money.

The custodian problem

A custodian holding crypto for institutions faces a choice with no good answer, and it comes down to how the wallets are arranged. It can keep everyone’s assets in one commingled wallet, an omnibus account, which is cheap to run and hides any single client inside the pool. Or it can give each client its own segregated wallet, which is what a client needs when doing anything on-chain, because a lending market or a DEX must see a discrete account before it will lend to or trade with that account.

The trouble is that a segregated wallet is a public track record. Every position the client takes, every loan, every collateral move, is tied to that one address and open to anyone. Watch it for a while, and you have the strategy.

Custodians keep client assets pooled in the omnibus. They move a client’s share out to its own address only when it needs to transact. That keeps the assets private while they do nothing. But the instant the client borrows or trades from that address, the activity becomes publicly visible. You can keep a balance private while it does nothing. You cannot keep it private while it works.

STRK20 will protect privacy even when assets are productive. You can now trade from the shielded pool without making the asset public. You will soon be able to borrow and lend privately, too. So, the behaviour stays hidden not only while the asset is idle but also when you put it to work.

Satoshi saw it coming

Bitcoin’s culture venerates sovereignty, yet its ledger is the most transparent. Pseudonymity hides who you are, only until one address is tied to you.

In a 2010 Bitcointalk thread, when a user raised the idea of using cryptography to hide transactions, Satoshi took it seriously before deciding that the tools of the day could not do it. “It is hard to think of how to apply zero-knowledge proofs in this case,” he wrote, because Bitcoin must prove a coin was not already spent, which seemed to require seeing everything. Sixteen years later, zero-knowledge proofs do exactly that, just not on Bitcoin.

Privacy matters more when using BTC. You can post wrapped bitcoin on Aave and borrow stablecoins against it without selling. The DeFi built around Bitcoin has grown from almost nothing two years ago to around $6.5 billion, with $7 billion in wrapped BTC. However, this is well under 1% of all the bitcoin in the world.

The primary reason for bitcoin’s limited use is the lack of programmability at the base layer. If you must use BTC, you must do so with a solution that relies on additional trust assumptions. That trust is the wall.

Starknet is going after the root of the problem. StarkWare is one of the main backers of OP_CAT, a proposal to re-enable an opcode Satoshi disabled in 2010, which would let Bitcoin verify STARK proofs directly and allow Starknet to settle on Bitcoin itself. It has already run a STARK proof through a verifier on Bitcoin’s test network and funded the research behind the push. If it lands, Bitcoin gets real programmability, lending, leverage and yield secured by Bitcoin rather than by a custodian. The catch is that OP_CAT is a soft fork, and Bitcoin changes its rules slowly and reluctantly. It is a multi-year bet, and it may never clear.

strkBTC is what works until OP_CAT is live. Yes, it is a wrapper with additional trust assumptions. But it is the wrapper that delivers DeFi with privacy, instead of waiting for a Bitcoin soft fork. We know how hard those changes to Bitcoin are to come by. Starknet is playing both tracks at once. The long game is making Bitcoin programmable at the base layer. The near-term aim is a private, usable version of bitcoin on Starknet, while the long game is being fought.

Compounding Privacy

Starknet’s design turns scale into a strength. Every new participant and every new asset in the pool makes it harder to single out those already inside it, so the system grows more private as it grows larger.

Safety in numbers

Chains used to compete on speed and fees. We have seen infrastructure rapidly commoditised. Transaction costs are racing toward zero everywhere. Liquidity is the only moat left. Hyperliquid is a live example of that.

A privacy pool is liquidity too. And it is harder to copy than regular liquidity. It gets more private as it gets bigger. A pool with only a handful of users gives them away through timing and matching amounts. A pool holding USDC, strkBTC, tokenised treasuries, and staking tokens, with thousands of people moving through it, makes it more difficult to recognise patterns. Every deposit is another plausible explanation for every other movement. Zcash is the proof in both directions. For eight years, barely a tenth of its supply was shielded, which is why researchers could trace so much of what passed through it. Then the crowd arrived, the shielded share climbed past 30%, and the pool grew harder to read. Privacy is a network effect, and unlike speed or fees, it compounds rather than commoditising.

The pool is most private to those who keep their activity within it. Because every time you cross the public-private boundary, some information can be used to construct your identity. So, the rational move is to stay in the pool. The move is rational for you and for other participants, too: for each user who stays, the next one becomes harder to single out. If Starknet gets this right, the privacy pool is the moat itself. Because a rival cannot match by simply shipping faster blocks or cheaper gas.

A lot must go right

Every privacy solution has had the cold-start problem. The pool is weakest on the day it opens, when the crowd is smallest, and the first user is the most exposed. Institutions will wait for more liquidity before they trust it. Privacy-natives may baulk at a compliance-heavy design. Builders wait for users, users wait for builders, and the pool stays thin. Zcash took the better part of a decade to fill, and it had a movement behind it. STRK20 must win the dull institutional flows instead, which move more slowly because they take longer to trust. It is entirely possible the pool never gets dense enough, and the whole thesis hangs on it.

The compliance design may not please anyone either. Privacy maximalists will read the viewing key as a backdoor and leave. Some regulators will decide it gives them too little, rendering it untouchable for institutions. Legal teams at banks may want precedent long before they want a novel design. The system risks being technically balanced and politically homeless at the same time.

The field is getting crowded. Aztec is building private execution from first principles. ZKsync has its Prividium model for enterprises. Canton already has the institutions in the room. Miden, spun out of Polygon, is taking privacy seriously. Starknet’s head start is its architecture, and crypto is full of better architectures that lost on distribution. It may also simply stay infrastructure. A private base layer produces nothing on its own. Someone must build a confidential settlement on it, ship private custody products, and wire it into lending markets and RWA platforms. While some pieces are already live, things like lending will be live soon.

A bug could do to STRK20 what one just did to Zcash’s Orchard. It is a shielded pool, so it carries the same risk. A zero-knowledge proof does not check that a transaction is honest, only that it follows the rules fixed in the circuit beforehand. When you omit a rule, a transaction that should have failed still satisfies all conditions, and the proof certifies it as valid. This is the most common way these systems fail, and it has nothing to do with the trusted setup.

Orchard didn’t have any trusted setup, and the bug happened anyway. Removing the ceremony, which is what STARKs do, closes one door and leaves this one open. Two things limit the damage. How hard the circuit is audited and how fast a flaw is caught, which is never truly finished. And whether the protocol caps what leaves a pool at what went in, so a note forged inside can never be withdrawn.

How Starknet finally brings institutions on-chain

Every privacy system before this made you choose. Keep your privacy and lose the market, or keep the market and lose your privacy the moment you touch the asset. Starknet does not force that choice onto you. An asset already on the chain goes private while continuing to trade with the same public liquidity and the same apps. There’s no other privacy design that offers a combination of no trusted setup, quantum-resistant proofs, settlement at protocol speed, and one shared liquidity pool.

We have been waiting for institutions, and they are here. But they still treat blockchains like back offices. Mainly as ledgers. Their assets are not yet fully programmable, as blockchains and smart contracts offer. Institutional assets come on-chain largely as representations of what happens off-chain and for settling those assets. Because it is cheaper. Settlement came on-chain. However, ‘putting assets to work’ stayed off-chain because using the asset in public discloses your position. A solution like Starknet closes the gap so that the asset can earn, lend, serve as collateral, and trade without its owners having to worry about surveillance.

Crypto spent years treating privacy as a way out of the financial system. It may turn out to be the thing that finally blurs the boundaries between the two.

Exploring how crypto expands traditional finance,
Saurabh Deshpande

This episode follows the money to the backbone of the AI boom, where David Choi of USD.AI is building the interest rate of intelligence, which is a debt market for the GPUs that everything else now runs on.

Spotify

iTunes

Banks have spent a century learning to lend against assets like houses, cars, office buildings, and even paintings. But they don’t give you a loan against one of the most valuable assets in the world right now: a GPU.

That is the gap David Choi is building into. He comes to chip financing from a background in lending against hard-to-price assets, from artwork to exotic real estate, at Deutsche Bank. The company started with his personal frustration when he couldn’t get a loan against his 300-unit server of H100s.

Why does this gap exist? Most assets became financeable because someone figured out how to turn the debt into something tradeable. Before Fannie Mae, you would roll over a home loan until the bank asked you to repay it. The fix was to rate it, package it, and sell it on. The primary lender is not stuck. That packaging machinery, the asset-backed securities market, is the engine that built modern finance. GPUs do not have one.

There are good reasons for that. A chip depreciates fast. It has a useful life of around six years, with newer generations landing every eighteen months or so. The traditional securitisation process, raising a warehouse facility from a bank, writing thousands of loans, going on a roadshow, tranching it, and selling it down, can take up to three years. That timeline is fine for a thirty-year mortgage. The value chain doesn’t know how to do it for an asset that becomes obsolete in six years. So banks pass, and neoclouds end up at private credit funds paying 20-25% all-in.

USD.AI compresses that machine onto a blockchain. A real-world use of blockchains, yay! The moment a loan is created, it is instantly folded into a single tradable instrument: sUSD.AI. Think of it as wrapped staked ETH, but for data centres. Because the debt is liquid, a lender is far more willing to fund you because you are not holding them hostage for years. The bet is to become the Fannie Mae for GPUs.

Instead of lending against the borrower, USD.AI lends against GPU chips. And no money is released until the hardware is installed and verified. The borrower does not own the chips while the loan is live, so USD.AI can take them back. The borrower has to keep a few months of payments in reserve. If the chips must be sold for less than the loan amount, Munich Re covers the shortfall. It helps that US data centres are metal boxes with high security and insurance, making them about the safest place to park a physical asset you intend to lend against.

The conversation spans several components of GPU financing, such as the mechanics of repossession, why Nvidia refuses to finance its own chips, how a gigawatt of power compares to fifteen Las Vegas Spheres, the trouble with buybacks as a token ritual, and why GPU-backed debt is not a niche.

Thinking about what backs the boom,
Saurabh Deshpande

In this episode, I sat down with Alex Gluchowski, CEO of Matter Labs, to ask why every major bank is now quietly building on a blockchain.

Spotify

iTunes

Banks do not trust each other.

That sounds obvious, but it is the most important reason blockchains matter to financial institutions. A bank trusts its own systems. It may trust its auditors. It may trust a counterparty after months of legal work. But it does not want to trust another bank’s server just because both are part of the same network. The whole point of blockchains is that they scale trust. That’s what banks want.

But what banks want comes with strings banks can’t accept.

For years, crypto people have framed the value of blockchains around decentralisation, public verifiability and permissionlessness. These are useful words from a technological standpoint, but they often make the conversation abstract. The simpler question is: can the rules of a financial system be enforced without trusting the people running the system?

That question becomes even more important when the assets are private.

If a public chain breaks, many people can see it. Traders, analysts, security firms, and rival teams are all watching the same ledger. A strange mint, a broken limit or an unusual movement in supply can trigger alarms because the data is visible. Public blockchains create a market for monitoring.

Private networks do not have that luxury. If the failure is hidden, it can silently spread.

This is the uncomfortable problem banks face as they move towards tokenisation. They want privacy, but they also want stronger guarantees. They want assets to move faster, but that can’t happen if they must build bilateral trust with all counterparties. They want the benefits of shared rails without turning the financial system into a giant group chat between bank servers.

Is there a solution that gives you the best of both worlds? Private execution with public-grade enforcement.

This is what ZKsync’s Prividium does. Tokenised deposits will be the starting point for most banks. Over time, this can expand into wholesale settlement, clearing, foreign exchange, securities, and more complex financial contracts.

The conversation explores the tension between what crypto offers and what institutions need. The episode lays out what makes a blockchain a blockchain, why public verification matters, where private networks fall short, how tokenised deposits may evolve, and why ZKsync is now betting on institutional finance.

Thinking about trust as infrastructure,
Saurabh Deshpande

Today’s issue extends our piece on vertical capital integrators. Blockchain rails are changing how value moves across the internet. But little has been written on the mechanisms by which traditional capital aggregators — Stripe chief among them — have expanded into crypto rails. This piece plugs that gap.

I break down how value leaks out of the payment stack at every step of a transfer, then map those losses to Stripe’s acquisitions over the past two years. Those acquisitions will add another $1 billion to Stripe’s topline by next year. We believe there will be businesses built, scaled, and acquired in each part of the stack Stripe has been going after, as the pie for internet-scale finance is constantly growing and ever-evolving. This piece should help build a mental model of the individual constituents of that stack.

As always, if you are building along these themes, find us on Telegram or drop a note to venture@decentralised.co. We’d love to chat.

In 2025, Stripe made $6.9 billion in revenue. It processed $1.9 trillion in payments. That’s about 6.5% of US GDP and 1.6% of global GDP. Over five million businesses use Stripe to accept payments, including 90% of the companies in the Dow Jones Industrial Average. By most measures, it is the largest internet payments company in the world. Among international businesses Stripe caters to, 30% of revenue comes from countries that are neither its home market nor among the top 10 global economies.

The concept of “domestic markets” doesn’t really make sense for internet businesses. The share of cross-border payments cannot be ignored anymore.

Global cross-border flows exceeded $190 trillion in 2024 and generated an estimated $240 billion in revenue for the banks and intermediaries that facilitate them. The average cost of sending $200 across borders is still 6.5%, more than double the UN Sustainable Development Goal target of 3%. That difference, between what cross-border payments cost today and what they should cost, is the single largest fee pool in global payments.

So Stripe doesn’t make much on one of its fastest-growing payment processing through correspondent banks, FX desks, and SWIFT messaging infrastructure that Stripe has no control over. Gaining better control over cross-border payments will allow Stripe to capture a much larger share of this growing pie.

Capturing a larger share without passing cost to the consumer means bypassing the existing intermediaries. That is what Stripe has been building. The only way to see what it is doing is to look at who the banks and financial institutions actually are, and what each of them is being paid for.

A cross-border payment is a relay race among seven functions, each collecting a toll before the money moves to the next.

Consider a merchant in Singapore accepting a $200 payment from a customer in Berlin. Stripe collects the transaction at acceptance and routes it at orchestration. A partner bank takes ~10 bps for licensing the cross-border move. A custodian holds the funds briefly. on the EUR to SGD conversion. A commercial bank or stablecoin issuer represents the dollars on the other side. SWIFT carries the messages. By the time $200 lands in Singapore, six different counterparties have been paid. Roughly $13 of the original $200 is gone before the merchant sees the receipt.

Seven layers, roughly in the order the money touches them:

1. Acceptance is where Stripe, Adyen, and PayPal capture payment from the customer. Take rates on a single cross-border card transaction are around 3% to 4%.

2. Orchestration is the API layer that routes the transaction and coordinates between providers. Stripe operates here alongside dedicated orchestrators like Spreedly and Primer. These players charge a few basis points on every transaction.

3. Licensing is how moving money across borders becomes compliant. If the operator does not have its own money transmission licenses, it must rent rails from someone who does. A typical US-to-Europe transaction routes through a partner bank or a licensed money transmitter that earns a cut for lending its regulatory coverage. Getting licensed independently takes roughly two years and up to $475,000 in filing fees, which is why most fintechs end up paying for access to someone else’s licenses.

4. Custody is where funds sit with a regulated entity while they are in flight. This layer is owned by custodian banks like BNY Mellon and State Street, which charge custody fees on balances and transaction fees on movements. Stablecoin issuers have their own Treasury custody relationships, which is how BlackRock and Fidelity ended up as major players in the stablecoin reserve market.

5. FX. Somewhere in the route, euros have to become dollars, and dollars have to become reais. Currency conversion is handled by FX desks at correspondent banks, which charge spreads of 10 to 50 basis points on institutional flows and substantially more on retail volume. FX is one of the single biggest line items in the $240 billion cross-border revenue pool.

6. Issuance is where the proxy for actual dollars comes into the picture. Stablecoin issuers allow companies and users to mint and redeem tokenised fiat currencies. Tether holds roughly $141 billion in short-duration Treasury reserves and earned over $10 billion in profit in 2025, almost all of it yield on reserves. Circle’s revenue on USDC works the same way.

7. Settlement is the final movement of value between institutions. In the traditional system, this is done over SWIFT messaging and a chain of correspondent banking relationships. This is where delays and costs compound. Stablecoin payments solve the delay by settling on a public blockchain like Ethereum, Solana, or Tron, but those chains charge their own transaction fees in their native tokens, which is money that leaves the payment stack entirely.

Every cross-border payment today passes through some combination of these seven gates. Most are owned by different entities. And that’s why it still costs ~6% to send $200 across a border. The $240 billion that banks and intermediaries collect on cross-border volume is the sum of what each layer charges.

Until 2024, Stripe owned only the first two gates. The other five belonged to banks, FX desks, licensed money transmitters, stablecoin issuers, and SWIFT. That is the value chain Stripe set out to replace.

Stripe acquired the next four layers with one Bridge Transaction

By 2024, Stripe had owned acceptance and orchestration for over a decade. When our customer paid for the software to the Singapore company, the euros entered Stripe’s Acceptance infrastructure in Berlin within milliseconds of the charge being authorised. The Orchestration logic performed necessary checks and handed the payment off.

In the past, that relay was where value leaked. A partner bank took a cut to facilitate the cross-border payment. Some FX desks took a fee for converting EUR to SGD. Another bank may have to hold the funds in custody while they wait for settlement. A stablecoin issuer, or more often a ledger inside a commercial bank, created the representation of those dollars on the destination side. SWIFT passed the messages that let everyone agree on what had happened. Each gatekeeper charged, and the charges were invisible to the Singapore merchant, who saw only Stripe’s fee on the receipt.

In October 2024, Stripe announced the acquisition of Bridge for $1.1 billion. The company had been founded only a couple of years earlier by two former Square executives. It was already being used by SpaceX to collect payments for Starlink in over a hundred countries whose banking systems couldn’t reliably process dollar transactions, and by Scale AI to pay data labellers in dozens of emerging markets. The acquisition looked rich against any revenue multiple. But it wasn’t their revenue that Stripe was going after anyway.

What Stripe was actually buying was a collection of regulatory and operational assets on the cross-border value chain. Bridge held money transmitter registrations across 30 states in the United States. It had custody arrangements with BlackRock, Fidelity, and Superstate for the Treasury bills that back stablecoin reserves. It ran an FX desk that marked the spread between fiat and stablecoin pairs. It later issued its own dollar-backed token, USDB. Stripe bought the next four layers in a single transaction.

By the spring of 2026, Stripe owned six of the seven gates. Acceptance and Orchestration had been there since the beginning. Licensing, Custody, FX, and Issuance were acquired with Bridge. The seventh gate was Settlement. That’s where Tempo comes into the picture.

By the first half of 2025, with Bridge integrated, Stripe controlled six of the seven pieces of the payment system value chain. The seventh was the rail on which a settled payment moved from one ledger to another. In the legacy system, that rail was SWIFT. A payment that cleared via SWIFT could take a long time to finalise and incur costs that compound with each hop along the way.

The obvious question was whether Stripe should use an existing blockchain as its settlement layer. Ethereum was the largest smart contract platform and had the deepest developer ecosystem, but at Stripe’s potential transaction volumes, base-layer fees, and block times were disqualifying for small payments. Ethereum does about 15 transactions per second.

Stripe processed 137,000 transactions per minute at peak during Black Friday and Cyber Monday in 2024, and about 2,300 per second sustained. and governance priorities, few of which were tailored to the operational requirements of a payments company. Solana was fast enough at 3,000 to 5,000 TPS, but it was a general-purpose chain with its own validator set

The alternative was to build a chain whose priorities were payment settlement. In September 2025, Stripe and Paradigm jointly announced Tempo, a payments-focused layer-one blockchain co-incubated by the two companies. Matt Huang, Paradigm’s co-founder, would lead Tempo while continuing to run Paradigm. A month later, Paradigm confirmed that the Ithaca team was joining Tempo to build the execution infrastructure. Paradigm’s own research group would stay on the architecture.

The Ithaca engineers were the core contributors to Reth, a Rust-based Ethereum execution client that Paradigm had been developing since 2022 and had shipped in production form in June 2024. Tempo required a skill set suited to a new chain designed around payment throughput rather than general-purpose computation at its foundation, and Ithaca provided that.

It made sense to build Tempo as EVM-compatible because the Ethereum ecosystem of contracts, libraries, and developer tooling had accumulated over a decade of engineering work that Tempo could inherit.

By mid-2025, the structural ingredients were in place. Stripe was in a position to completely own the stack.

Tempo’s mainnet went live on 18 March 2026 — eighteen months after the Bridge announcement and six months after the Ithaca team formally joined. The chain shipped with the architectural decisions that its payments-first design required. It ensured three critical features:

1. Gas fees are paid in approved stablecoins rather than a native token, so merchants never have to hold a second, volatile asset to move payments on the chain.

2. Sub-second deterministic finality means a transaction is cryptographically final at block inclusion, faster than any existing cross-border settlement rail and close to what card networks deliver at checkout.

3. Every transaction costs a tenth of a cent per transfer, cheap enough for cross-border payments to undercut ACH and for agent-to-agent micropayments to survive their own fee.

In April 2026, Visa joined Tempo as an anchor validator after six months of joint engineering with the Tempo team, taking its place alongside Stripe and Zodia Custody, Standard Chartered’s digital-asset subsidiary. Visa’s cross-border settlement business runs on correspondent-bank rails that Tempo is designed to bypass, which made its decision to operate a validator node a tell about where the industry expects this to go.

The stack closed in April 2026. For our $200 from Berlin to Singapore, every fee on every layer was now Stripe’s.

A billion in additional revenue

If the Bridge-Tempo stack can achieve a monthly target of $10 billion a month, it will generate roughly $1 billion a year in additional recurring revenue, on top of the merchant fees Stripe already collects on cross-border traffic. The ceiling is substantially higher than the 2027 target; Tether generated more than $10 billion in net profit in 2025 on roughly $175 billion of USDT in circulation, and that was the take from one layer of the seven Stripe now controls.

Here is how Stripe can make $1 billion with its vertical integration.

Acceptance and Orchestration are Stripe’s existing merchant fees on cross-border transactions and the baseline that the new stack protects. Nothing new here.

Licensing keeps $100 to $200 million a year inside Stripe. All the licences that Bridge acquired carry the regulatory coverage for cross-border dollar flows directly, eliminating the partner-bank licensing fees Stripe would otherwise pay on that volume.

Custody is a small cost line on the reserve balance. Bridge retains the Treasury yield that flows through BlackRock, Fidelity, and Superstate.

FX earns Bridge a ten-to-fifty basis-point spread on every stablecoin-to-fiat conversion. At $120 billion of annual volume and an average 30-basis-point spread, the FX layer generates roughly $360 million a year.

Issuance earns the reserve yield on USDB’s Treasury backing, roughly 4% to 5% in the current rate environment. At the 2027 volume target, an expected float of roughly $10 billion contributes about $450 million. Tether’s $10 billion of profit came on roughly $175 billion of USDT.

Settlement earns sub-cent per-transaction fees on Tempo and is immaterial as a revenue line at any realistic volume.

Across Licensing, FX, and Issuance, the 2027 target amounts to roughly $1 billion in potential annual revenue. That’s about 15% additional revenue every year. As the supply of USDB grows, Stripe earns yield on it just like Circle and Tether do on USDC and USDT backing, respectively. This becomes additional revenue for Stripe.

The stack describes a world where people pay for things. But we are also entering into a time where agents will pay for things. Stripe is building a second stack where software pays for things.

Gartner forecasts that AI agents will intermediate $15 trillion in global B2B spending by 2028. McKinsey expects agentic commerce to orchestrate roughly $3 to $5 trillion by 2030, while Bain estimates that US agentic commerce could reach $300 to $500 billion in the same period.

What is actually settling on agentic rails today is closer to a rounding error. Bloomberg reports that Coinbase’s x402 processed about $24 million across 94,000 buyers and 22,000 sellers in a single 30-day window in early 2026. The point is that the gap between estimates and what’s happening now is massive. If the forecasts even partially come true, companies that get a take rate on every dollar moved by agents will win. If they don’t, Stripe still owns the merchant network and the cross-border revenue stack it already monetises today.

Stripe has been assembling the pieces of the agent stack on different rails since before Tempo existed as a product.

The wallet layer is Privy, the infrastructure company Stripe acquired in June 2025. At the time of the deal, Privy powered more than 75 million accounts across over 1,000 developer teams, orchestrating billions in transactions. Privy’s embedded wallets live inside the applications developers are already building, with login via email and keys split across secure enclaves so no seed phrase ever touches a user. Every agent spun up against a Privy-backed application inherits a wallet at creation. The wallet becomes a default rather than a setup step.

The rails are the Machine Payments Protocol (MPP), launched alongside Tempo’s mainnet in March 2026. MPP is a specification, co-authored by Stripe and Tempo, for how software makes requests and payments. It resurrects HTTP 402, a status code defined in the original HTTP specification as “Payment Required” and left dormant for nearly 30 years.

MPP can do everything x402 does, and more. It can handle simple per-request payments, but also supports sessions, where many small payments are grouped and settled once. x402 doesn’t support that. That means MPP can work with x402-style services without changes, but not the other way around.

Within its first week, MPP was integrated across more than 50 services, including OpenAI, Anthropic, Google Gemini, and Dune Analytics on the AI side, and purpose-built agent-commerce services like Browserbase (pay-per-session headless browsers), PostalForm (programmatic physical mail), and Parallel Web Systems (per-call web access). Visa contributed to the MPP specification alongside Stripe and Tempo.

Stripe’s stack is the first end-to-end position in that world. Privy issues the wallet. MPP structures the request. Tempo settles it. USDB carries the value. Stripe remains the merchant of record, earning its acceptance fee on a rail that costs a fraction of a cent to operate. If Gartner’s directional number is even half right, the ceiling on this stack is well above the cross-border merchant volume.

Ethereum, Tron, and Solana moved the trillions of dollars of stablecoin volume. Then why are companies like Stripe trying to launch their own chains and hoping that they will attract enough volume? About 15%-20% of the current stablecoin volume can be considered non-speculative as per Visa. The share of stablecoin use in daily commerce, i.e., the non-speculative share, can only grow when stablecoin rails are integrated into merchant networks. It is difficult for general-purpose chains to do that.

This is where the idea of Tempo, Arc, and Plasma starts to make sense. The argument for the new chains does not rest on Ethereum or Tron breaking down. General-purpose chains are at the bottom of the entire stack. Besides settlement, six other functions need to be in place to complete the value chain. And it will happen for whoever gets there first.

Tempo is not the only operator making this bet. Arc and Plasma are parallel attempts with different parts of the stack already in hand.

While payment chains are optimised to capture more of the traditional settlement business from merchant flows, existing chains are far ahead in financial speculation use cases. The volume of chains like Tron, Solana, and Ethereum is orders of magnitude higher than payment-specific chains right now. These chains also offer much more liquidity. But the bet is that the use of stablecoins for so-called genuine businesses will be as much, if not more.

The following image shows how these new chains, especially Tempo, are geared towards the payments use case.

The six chains fall into three categories. Tempo and Arc are operator-controlled stacks, owned by the commercial entity that also owns the issuance and the distribution. Plasma is purpose-built for stablecoin payments but independent of any issuer or merchant network. Ethereum, Tron, and Solana are the incumbent public chains hosting the bulk of today’s stablecoin volume. The public chains, for all their current dominance in volume, capture only gas.

Every layer except acceptance and orchestration is capital-intensive but buyable. Bridge cost $1.1 billion and delivered issuance, FX, and licensing in a single transaction. Privy was a smaller check and delivered 75 million embedded wallets. Ithaca was an acqui-hire. Spinning up a new chain like Tempo is not a big deal these days. MPP is an open specification that anyone can clone. Five of the seven things on the stablecoin payments stack can be assembled for a few billion dollars and a two- to three-year roadmap.

Acceptance and orchestration are the points where you interface with the customer. You can rent the highly desirable access to these layers. You will take a long time to clone them. You can issue stablecoins like Circle and Tether or build a new chain like Plasma. But you cannot buy into the checkout flow of five million merchants. Adyen has scale but a different developer footprint. PayPal has consumers rather than a merchant API relationship. Stripe is the landlord now, and every stablecoin issuer that wants to reach commerce has to negotiate rent with the landlord or build its own checkout flow.

What Stripe built over fifteen years in checkout integrations and developer trust is the asset that the image does not show. The stablecoin stack is the monetisation wrapper on top of it. Bridge, Tempo, Privy, and MPP are tools that turn an existing moat into rent on every new transaction type, including agent payment volume that does not yet exist.

Circle can match Tempo, USDB, Privy, and MPP given enough time and engineering. What Circle cannot match, and what Tether cannot match, is Stripe’s merchant network. Everything else in this story is Stripe installing toll booths on a highway it already owns.

The last mile is the moat

PayPal launched in 1998. Wise in 2011. Revolut in 2015. Every few years, a digital remittance company arrived with better rates, better UX, and we felt Western Union was done for sure. But not yet. If you go to Argentina, the best way to swap your dollars for pesos is Western Union. It moved roughly $76 billion in consumer-to-consumer transfers last year, and its network runs to 500,000+ agent locations across more than 200 countries.

Look past the agents. The real asset is the postal system.

Western Union is wired into the post offices of more than 100 countries. It rents the last mile from post offices of different countries.

Visa did the same in 1958, when BankAmericard piggybacked on the branch networks of every bank willing to sign. Safaricom ran the same playbook with M-Pesa, turning airtime resellers into cash-in and cash-out agents and pushing ~$300 billion a year in transaction volume through a telecom’s distribution. Alipay did the same by living inside Taobao. Tencent won payments without a single bank branch, because WeChat Pay was integrated into a messenger with 1.3 billion monthly users.

Every one of those landlords had a statutory duty to rent. Post offices are public infrastructure with an obligation to let anyone inside who can pay the counter fee. Bank branches under the old correspondent regime had the same obligation as Airtime agents, who were licensed under telecom rules that forced open access. WeChat is the partial exception, and Tencent spent a decade being pressured by the PBOC to keep the rails interoperable. The post offices worked because the doors were held open by law.

Stripe’s doors are not.

Now look at what Stripe is doing with Tempo.

Stripe doesn’t need Tempo to find merchants. Stripe already has them. Its 2025 annual letter reports more than 5 million businesses on the platform, $1.9 trillion in total payment volume for the year, 90% of the Dow Jones, 80% of the Nasdaq 100, and a quarter of all new Delaware corporations signing up through Stripe Atlas. That is the substrate. Tempo is what runs underneath it.

The pitch is straightforward. If you are already on Stripe, the chain is already there. Stablecoin settlement in real time, near-zero fees, and none of the 200-basis-point tax that Visa and Mastercard have been collecting for decades. What the merchant sees is a cheaper invoice and a cleaner reconciliation file. They don’t know or care whether the chain is used for settlement. If it’s faster and cheaper on existing rails, it gets a thumbs-up from the merchants.

While Stripe has the distribution for the payment use case, general-purpose chains rule the speculation use case.

Ethereum, Solana, and Tron carry real stablecoin volume. Tron alone clears more than $20 billion of USDT on an average day. This volume lives and dies with wallets held by traders, OTC desks, and remittance corridors. Public chains own the trading layer, and hold roughly 42% of all circulating USDT

The commerce layer sits behind Stripe, Adyen, and a dozen PSPs whose merchant contracts were signed a decade ago.

Plasma tried to shortcut the problem by buying liquidity directly. The TVL dropped from the peak of $6.3 billion to ~$1 billion as the incentives dropped, as the price of the XPL token dropped from ~$2 to $0.10. The thing is, you cannot bribe your way into a merchant integration. That takes years, licences, enterprise sales teams, and the kind of contracts that do not get signed on Twitter.

Arc has Circle behind it, which buys it USDC distribution across every major exchange and a growing share of fintech treasury stacks. Useful, but being in Coinbase is not the same as being inside the checkout flow of a US SaaS business billing a European enterprise customer. Stripe is the one sitting there.

The existing merchant network is the worldwide network of post offices. Merchants are vendor-locked into their payment service provider. They don’t care which chain their vendor points to. As long as the terms are fair, they will keep using the same vendor, whether the chain is Solana or Tempo.

There are two broad use cases of stablecoins. Financial speculation and commerce. Public blockchains have dominated and will dominate the first part. For the second part, we need existing merchants and existing payment flows. Existing companies like Stripe are in a much better position to capture this market.

Stablecoin chains by themselves don’t make sense unless you already have an existing volume. Owning a stablecoin chain doesn’t guarantee that you will make money. The stablecoin chain will be a loss leader. It lets you own payments end-to-end, so that once you own the customer, you can earn money from other services built on top of it. The moral of the story is that Stripe doesn’t need Tempo to win commerce. It needs Tempo so that it doesn’t have to share.

Watching fintech coming for crypto,
Saurabh Deshpande

Note - I use the terms supply side aggregator, demand side aggregator and vertically integrated capital ecosystem quite frequently in this piece. For ease:

1. A supply side aggregator is one that pulls in differentiated market participants to offer a commoditised product (eg., Uber).

2. A demand side aggregator is one that expands offerings to a collective that looks similar from the outside (eg., Amazon).

3. A vertically integrated capital aggregator is a financial entity that brings together multiple parts of an ecosystem to offer users several products within the same venue.

TL;DR - Blockchains are money rails. The value of a protocol lies in its economic output. Composability and real-time verifiability allow blockchain native businesses to integrate vertically. Tokens allow participants across the stack to be incentivised with a shared medium. Teams that are intentional about how each part of the stack accrues value, have moats. Vertical integrations aid in accelerating capital velocity within ecosystems. Done well, they become a source for revenue.

The short, twitter version of the article is linked here

This is a precursor to a series exploring the slow, but gradual enterprise-ification of crypto.

It is a beautiful Saturday morning in Bangalore. As is often a habit, I spend most of it in Cubbon Park. Built over a century ago, the trees there are often older than the country they find themselves in. Near it are old buildings that were once used to design fighter jets. “2607,” I say for my OTP as the driver plugs into the streams of traffic that choke the city. I rush for Spotify to drown out the noise. J. Cole’s latest album to the rescue.

Boredom strikes a few moments later, and I seek refuge in Substack. Maybe I could order snacks from Swiggy - the equivalent of Uber Eats in town on my way home.

In a matter of minutes, I’d have used four aggregators, each of differing kinds.

Uber aggregates riders across town and positions itself as a demand aggregator. So does Swiggy. You may not find much difference in the average rider or hungry person. They can be commoditised. So as a platform, Uber gets to take 30% of each commodity (I mean human users) they have aggregated to scale. The restaurants hate this. So do the drivers. Both try to circumvent by asking the user if they can pay directly in cash and skip the platform.

The rider (or me, in this case) understands that the platform’s ability to enforce a reputation system is what makes it valuable. No cash transactions today.

You see a version of this with Jupiter on Solana. Its influence in its early days was a byproduct of its ability to route orders across exchanges and offer the best prices to users. Instead of restaurants and drivers, Jupiter aggregates venues for my WIF buy orders to flow towards.

Substack and Spotify, however, operate on different parallels. Spotify pays up to 70% of its revenue to labels. As of last year, it paid roughly $13.75 billion to rights holders out of $20.22 billion in revenue. $0.04 of each dollar it makes reaches the artist. Substack gets to keep just 10% of what I may have paid the author of the newsletter I was reading. It has neither commodified the reader nor the writer. It simply positions itself as a tool. Perhaps, intentionally because if it were to command pricing power, the writers on the platform may simply abandon it.

Spotify is an instance of a loosely knit supply side aggregator by virtue of the number of record labels it has collaborated with. Substack on the other hand positions itself as a demand side aggregator with no pricing power by virtue of the readers on it.

Each of these apps aggregates a different side of the market. But their ability to accumulate capital (or pricing power) hinges on how well knit they are within their ecosystem. We will soon see a different version of this.

Web2 native aggregators tapped into an abundance of users that came online through two forces.

1. First was Moore’s law working its magic on the cost of a smartphone. India alone has 800 million smartphone users. Roughly 5.5 billion individuals are online today. (Cheap phones, yay!)

2. Reduction in internet costs and the rapid decline in cost of bandwidth. (Cheap internet too! yay!?!)

Crypto as an economy - in contrast has a much smaller TAM. By best estimates, ~560 million users have interacted with crypto. Last month, DeFi saw ~10 million active wallets. These are distinctly different economies, with very different rulesets.

One runs on attention. The other, powers through what flows through your wallet on-chain. We often convolute behaviour from the attention economy into what may happen in the transaction economy. People on prediction markets for instance do not behave the same way they do on Instagram.

Revisiting Aggregation

Three years back, when I first wrote of aggregators, I argued that blockchains reduce the cost of verification and trust. The original promise of the web was access. You could sit in your pajamas in New York and buy Temu goods from Shenzhen. I argued that blockchains make it possible to verify and settle a transaction with the vendor in real time at minimal cost. Or so I prophesied the industry would evolve.

Writing in 2022, with my first sponsor for this newsletter (thanks Nansen) I wrote:

We believe that blockchains will enable an entirely new category of markets that are able to verify on-chain events instantaneously. This will collapse the cost of validating intellectual property at a massive scale and thereby create new business models.

In the years since, the new markets did come on-chain. NFTs did ~$100 billion in volume since. Perps became an economy that cleared ~$14.6 trillion since. Decentralised exchanges cleared ~$10.8 trillion. The understanding that the technology will be used to clear transactions with anonymous counterparties at a global scale was true.

But at the same time, it ought to be noted that OpenSea’s volume went from ~$5 billion back then to ~$70 million this month.

Markets emerged, evolved and died. As with many things in life. In the process, they left us fuel to meander and ponder upon. Sid believes that the speculation in crypto is a feature not a bug. All new markets are undefined in the early days with participants not understanding what they are buying or why something may hold value. The novelty is baked into the price. As sanity returns, valuations become efficient and the bubble becomes memories of times past. NFTs and DeFi have had their own cycles of such mania.

These bubbles were crucial to stress-testing and validating the monetary rails that eventually set the stage for markets like the one on Hyperliquid.

Which leads me to the past few weeks of operating in crypto. Between Drift and Kelp, close to ~$578 million was hacked in the past three weeks. In the trailing twelve months, roughly ~$1.7 billion was stolen across DeFi and crypto protocols tracked by DefiLlama. Excluding stablecoin issuers - all of DeFi made roughly $3.42 billion in protocol revenue over the same window.

In other words, this past year, close to $0.50 was lost to hacks for every $1 DeFi made in revenue.

At the same time we are seeing an increase in the number of apps being released. Software itself has become a commodity. The App Store saw ~84% more apps being submitted this quarter compared to the same quarter last year. We are simultaneously seeing two forces at play. More software demanding the attention of fewer users. And fewer platforms dominating the bulk of the revenue created within crypto.

A moment back, we were with the trees in Cubbon Park. We are now realizing we lose $0.50 on every dollar earned in the industry. Sharp curve I know - but stay with me.

If we break down the $3.5 billion earned across decentralised avenues, you will see a quick pattern. Roughly 40% of it comes from derivatives platforms. Hyperliquid alone contributed some ~$902 million towards it. The second largest category is decentralised exchanges, within which the leader Uniswap made some ~$927 million in fees. Lending platforms like MakerDAO follow at third, with close to ~$500 million in combined revenue. What is common amongst all of them is that they are capital dense businesses.

Unlike products that can be built with lines of code on a Saturday afternoon, they require coordination of patient capital that is willing to accept the risks of these platforms. I explore these themes in great depth in the two stories linked below.

This is where you realise the primary difference between web2 aggregators and web3 native ones. Because blockchains are primarily toolsets to move money and verify that the nature of the transaction follows rule-sets created by developers - they become valuable only when they are able to engage in capital dense operations. Perpetual exchanges are able to put large sums of money to productive use repeatedly on the same day. Lending platforms skim a small share of the large amount of money generated as yield.

Aave for instance earned ~$123 million off the ~$920 million it made in yield over the past year. But such aggregators dominate only when they are able to own three key things.

1. The supply side (of liquidity),

2. The demand side (of users) and

3. Distribution.

Hyperliquid is a unique beast in this regard. It has paid out close to $100 million in builder code fees but the vast majority of revenue comes from its own native front end. It is simultaneously able to retain its best users while expanding the surface area for new ones to interact with the protocol.

But what is the logic behind that? One theory is that distribution is a toll-fee paid in Web3. Large protocols tend to own and retain their best users. You realise this when you notice what amount in revenue has been created by decentralised exchanges, compared to on-chain aggregators who route orders.

On Ethereum, aggregators account for 36% of all DEX trades. On Solana, that number can be as low as ~7%, depending on the month. Kyber, 1inch, CoW and ParaSwap have combined made about ~$112 million in fees since launch. Compare that with Uniswap’s ~$5.5 billion in lifetime fees as the standalone exchange that sees the bulk of the volume. You can see a similar phenomenon in Hyperliquid too.

Combined, builder codes account only for ~6% of the cumulative ~$1.1 billion Hyperliquid has made in revenue. MetaMask’s tight-knit integration on Ethereum has earned it $184 million in fees last year through swaps. Phantom generated close to $180 million but consider that this is a small chunk of a huge economy of scale ecosystem. These products work when they are built atop a single protocol that has both the liquidity and the economic activity that is required to drive value towards them.

These products are able to attract and retain users because of the depth of liquidity available on them. Look through these lenses - capital in crypto is no longer a commoditised product. It is the most necessary ingredient. Vertical integration of capital gives participants more reasons to stay within the ecosystem. In such systems, capital begets liquidity because it can be put to productive use.

Capital is not the moat. It is a consequence of vertical integration you built. Vertical integration is the moat, and capital is the byproduct

The caveat is that it works only when parking capital is not incentivised. Don’t believe me? Take any pre-launch protocol that has an airdrop programme for deposits and you will see the case in point. Or the countless L2s that have struggled to generate value.

Any business that does capital aggregation is by extension a hack target. Drift was a target because it had ~$570 million in TVL. KelpDAO was a target because it held close to ~$1.6 billion in restaked ETH. Hyperliquid’s bridge holds close to ~$2 billion in user deposits, making it one of the most valuable standing targets in the space. You could see this play out with Ronin (~$625 million) and Nomad (~$190 million) too.

Because blockchain native businesses need high amounts of capital to be valuable, we have a dynamic where to win requires being vulnerable at least until the mechanisms to secure capital and freeze capital movements are in place.

And even when you do have capital, having a large TVL does not warrant success on its own. Capital in an economy that is parked idly, or put to little use can be a liability in the event of a hack. This is why you see protocols trying to differentiate themselves on the basis of what economic output they are able to generate, from niche usecases.

1. CHIP, the firm behind USDAI, has been able to place close to ~$100 million in loans this quarter, with a $1.5 billion pipeline in motion that will generate close to ~16% APY on its riskier tranches over the course of the year.

2. The highest risk pools on Maple generate ~15–20% APY which is comparable to or higher than the ~12.6% APY on USDC pools on Aave today. It has aggregated borrowers that can use the liquidity on the protocol to create that economic output.

3. Naturally, Hyperliquid is the best instance of a supply side aggregator that has been able to put capital to meaningful use. The last year saw Hyperliquid do ~$942 million in revenue with an average of roughly ~$3.5 billion in TVL. By some very crude math, each dollar parked on the protocol has rotated close to ~285 times annually, yielding ~$0.30 in fees per $1 of TVL. Compare that to Aave’s ~$0.05 in fees generated per $1 of TVL across its lending markets.
   

I can almost hear my finance professor screaming at me for comparing the yield on TVL between Hyperliquid (an exchange protocol) and Aave (a lending avenue). But in a market where consumer preferences are not predetermined, and investor loyalty is not deep, capital will flow where it produces the best outcome. When accounting for hacks, investors will demand a premium for the risks involved. As it stands, perpetual exchanges are the one place that are able to take idle capital and repeatedly use it on-chain to produce fees.

I initially thought Hyperliquid is simply a supply side aggregator. It supplies capital for users willing to trade on-chain. This has been my thesis. And that is indeed true. But it breaks when you consider how it uses tokens to incentivise vertically integrating into it. But before we go there, let me explain how a vertically integrated ecosystem works.

Ticketmaster is responsible for 70% of all major live events in the United States. It is able to take 30% off tickets you buy to perhaps see Justin Bieber play his old tracks on YouTube at Coachella because it owns the venue, promotes the tour, ensures you have merch at the concert and coordinates with sponsors. The 30% you pay, is 15 times what Stripe would charge to process a ticket sold online. But you pay that premium because Ticketmaster is vertically integrated across the value chain.

You have the illusion of a marketplace: artists, venues and fans are all parties but nobody gets to argue with Ticketmaster on its cut. The same applies to Apple’s App Store. Apple is responsible for curating the store, collecting payments, ensuring the device works and bringing in the millions of users that are now attuned to the ding Apple Pay makes when you subscribe to yet another app you will never use.

Vertical Integrations in Crypto

Protocols have slowly begun carrying out the same thesis.

In Web3, a capital provider can be treated as commodity if vertical integrations do not make it easier to work with. Users do not have loyalty until the ecosystem is integrated such that the cumulative experience is non-existent anywhere else.

For Maple, such an integration would involve years of experience working with hedge funds and market makers. For Centrifuge, the integration involves receiving close to ~$1 billion from Grove for their JAAA bond offerings from Janus Henderson. Instead of capturing loose, abstract parts of an economy, they integrate vertically to offer the end-user a better product. In doing so, they create moats that are hard to replicate overnight.

Maple’s years of underwriting experience, or Centrifuge’s moat in being a trusted capital coordinator are moats in a world where capital and relationships have become the only things that are hard to replicate.

Firms doing vertical aggregation may routinely give up parts of the stack to a third party. Part of the reason for doing it is the lack of economic benefit from doing so. Maple owning custody or Metamask issuing its own cards may not be massively profitable when compared to the capital they generate in swaps and credit underwriting.

However, when a business scales exponentially, owning the whole stack is where the competitive edge is built. This is partly the reasoning behind M&As in the industry.

When a business has a vertical integration in play, you are not competing with a single product. The battle is against the combined experience a user has within it. On Hyperliquid, once HIP-4 is on, a user would be able to be on-ramped for free (through Native Markets), take a prediction market position, and use that position as collateral to trade on its perpetual product. Its risk engine allows this experience. And for what it’s worth, this is not possible even in TradFi today without a merchant banker.

Hyperliquid owns the user, on-ramp, risk engine, trade interface, liquidity and issues the token. For a new business to compete against it, would mean battling on six different fronts, all at the same time.

For new apps that are launching, tapping into a small share of this is meaningfully a better outcome than building on a new protocol like Monad, which has done about ~$2.6 billion in cumulative derivatives volume across all five of its perps protocols.

Integrated ecosystems like the one on Hyperliquid attract developers, more integrations, headlines and happy token holders. (Or in my case - a writer that spends his time obsessing about how these things work)

Exchanges see this shift. Coinbase acquired Deribit, has a custodial business, co-issues the currency (USDC) with Circle and captures revenue on reserves, a massive wallet infrastructure and on-boarding rails in 100+ countries. It also launched its own chain in pursuit of vertically integrating the experience. Admittedly, Coinbase may have been too early in pursuing retail users that clearly did not want to “coin” content on a blockchain or use Farcaster.

The integrations for Coinbase exists in loose terms but sits hidden behind layers of bureaucracy, regulatory hurdles and internal priorities. This is perhaps the primary difference between an open-access integrated system and a walled one. Coinbase by virtue of being an exchange with ~$60 billion in market cap has little incentive to pursue the marginal developer.

Hyperliquid in contrast benefits from developing its core avenue to the best place to trade and simultaneously creating an ecosystem whilst driving value to the underlying token.

The token is part of the integration in this context as it is the shared substrate that keeps these integrations alive and valuable. This is where you see the industry confusing itself between a tokenised protocol, and a tokenised business. A tokenised protocol is predicated on the ease with which a third-party developer can build on it. It incentivises people to drive value (downwards) to the token - often in the form of buying it back from the market.

Firms like Robinhood and Coinbase are powerful economic participants, but they cannot replicate Hyperliquid’s core network of owner-operators.

The protocol’s airdrop ensured the people that own it, were individuals that contribute economically towards it. They own enough tokens to warrant driving value towards it. Hyperliquid commits to that cause, by giving up 99% of its revenue to buy the tokens back from the market. Imagine a listed firm using 100% of its revenue to buy back ESOPs from its employees. We may see people become more accepting of capitalism perhaps.

And this is where you see how the industry is evolving whether we like it culturally or not. Solana focuses on immutability. Ethereum has its focus on censorship resistance and open source. But observe carefully where traction resides and you will see the industry adjusting ideology to commercial reality.

Hyperliquid, while a beautiful garden, is a walled one. The source code is not open-source as far as I understand. How its risk engine works is not verifiable either as far as I understand. Nobody knows what parameters go into Maple’s risk underwriting. (Or at least, I don’t). As a lender, I may not even know who underwrote the loans on USDAI. But the question is, are these parameters being openly accessible a requirement for me? I don’t quite know. Probably not.

Negotiating With Chaos

An economy cannot be built if half a dollar on every dollar created in revenue is lost to hacks in a single quarter. Founders will flock to AI, if they are expected to be responsible for hundreds of millions of dollars parked as TVL on their product. Each time there is a hack we rush to hope stablecoins are frozen. Which in turn, are often centralised.

A vertically integrated stack will eventually require giving up complete decentralisation for the sake of economic progress.

And this is not a new story on the web. In the late 1990s, visions of an open internet that allowed free speech without consequences were dreamt of heavily. Nazi memorabilia was once auctioned on Yahoo until French courts intervened in 2000. Tim Wu explores this theme in great depth in Who Controls the Internet? The story of the web, or perhaps all human commercial networks - is that of complete decentralisation paving way for a toned down version that gives up controls for the sake of economic interaction.

We settle for a diluted version of the original so that the commerce can scale, because in the absence of dilution, chaos will ensue.

This massive expansion in energy is caught in how we describe those ages. The “wild” West. The dot-com “boom”. Perhaps, even crypto is undergoing a similar expansion and toning down of energy. I explored these themes in great depth last year in the stories below.

What does this mean for founders? Observe Metamask and Phantom’s numbers. Those businesses make more money than most L2s, because they are downstream in an ecosystem with tremendous economic output. Building bridges to nowhere and exchanges for nobody is no longer a business model. Especially when the pain associated with it involves hacks. You build where the liquidity and users are today.

Mimicking a vertically integrated product overnight may not be possible. But you can build atop it.

Operating systems saw a similar pattern. When Blackberry declined and iOS became a dominant player - developers had to choose where they built. We are seeing a version of that in crypto too. Except this time, capital incentives may keep developers blinded longer.

Remember how I mentioned Cubbon Park? I often think of why the park exists clean and well kept for the decade or so I have spent time in Bangalore. Part of the reason is that there are rules. You are not allowed to spend time in it past 6 pm for instance. Or hosting an event in the middle of it may get you screamed at. Sometimes, they are harsh. But the rules keep the place functional.

Platforms and protocols on the internet are a lot similar. We may not like rules around it but they keep things functional.

In the age of vertical integrators, we may find ourselves increasingly agreeing to rules so that our dollars stay with us and the economies we spend so much of our time on, continue to scale. The trends point towards this. Stablecoins, RWAs, perpetual exchanges with closed-access risk engines, lending desks with risk underwriters that are unknown and off-chain RFQ products like Derive all point to the same trend.

That of vertically integrated capital aggregators that are willing to give up dreams of complete decentralisation for the sake of progress.

Reading loads of Neil Postman
Joel John

TL;DR: Decentralised.co is increasing its exposure to perpetual exchanges through a position in Drift Protocol. Solana is yet to benefit from positive tailwinds in the growth of perpetual exchanges. Drift is best positioned to benefit from RWA assets and internet capital market instruments issued on Solana. At a price-to-earnings multiple of five based on last year’s earnings, it is deeply discounted relative to its position in perpetual exchanges.

This brief note should help explain why we are going long on the long-standing perpetual exchange in Solana. Venture@decentralised.co is the best place to reach out if you are building in the theme, and would like to chat.

Over the past month, we built a position in Drift. The protocol generated $19.8 million in fees over the last year. It currently trades at an FDV of $95 million. At the same time, it sits at a price-to-book of 3.5x on its operational balance sheet alone. We believe Solana will continue to be the chain on which fintech applications are built and scaled. Drift is the longest-standing perpetual exchange within it. These two forces blend to make it an interesting asset to own. Our primary reasoning for involvement with Drift boils down to three core factors.

1. Drift’s team has worked relentlessly towards scaling perpetuals on Solana. We are comfortable going long on the team itself.

2. Drift’s architecture for providing liquidity will become valuable in a world where centralised sequencers are seen with suspicion. DLOBs, JIT, and vAMM collectively offer a better experience to end users. Open, verifiable markets will be a necessity as on-chain markets evolve.

3. The DeFi landscape on Solana has significant runway for growth in equities, real estate assets, and internet-native capital instruments. A perpetual exchange will be needed for sophisticated actors to interact with these instruments. Drift can be the preferred avenue for price discovery and hedging.

To keep terminologies tight, we use the following throughout this note.

• Earnings - What is left in the protocol’s balance sheet after cost of incentivising economic activity

• Revenue - Total dollar value made by protocol and users

As a general heuristic, we use price-to-earnings throughout much of the document, except in the initial section.

Before getting into the why for Drift, it helps to understand how its fundamentals stack up.

Drift Protocol is the longest-standing perpetual exchange on Solana. It currently holds a quarter of all TVL allocated to perpetual exchanges on Solana. Its yearly revenue grew by 193% in 2025, from $14 million to $41 million. Annualised, in 2026, the protocol is on pace to generate $30.75 million in revenue, extrapolating from what the protocol made in January and February. (Revenue here combines fees paid by users alongside staking rewards from Drift-staked-Sol). Drift holds $27 million in its fee treasury accumulated entirely from protocol earnings over four years, excluding the $52.3 million raised from investors.

Perpetual exchanges are booming as a category, but they are not entirely open, verifiable, or decentralised. Part of our interest in Drift comes from its open-source nature and direct settlement on Solana. But before we go into how it operates, it helps to have a perspective on why open, verifiable financial networks matter.

Open Markets Need Verifiable Infrastructure

Human institutions work well during periods without stress. A dictatorship can be more efficient than a democracy until the citizenry questions those in power. People think about the mechanism that elects someone to power only when the power is abused. The same applies to markets. We don’t often think about how a market functions until something breaks. Markets, like democracies, are bundles of rules that must function in sequence. When the rules are not followed or broken, trust in the market declines. And a byproduct of that is citizens (or traders) simply moving elsewhere.

This is why trustlessness is a fundamental need for markets to evolve. Bitcoin serves this function for spot assets. AMM model markets on Ethereum and Solana solve this for spot markets. The fact that we know how much money is lost to MEV is a feature that happens because the rulesets and infrastructure through which a spot order is settled on-chain are verifiable. Perpetual exchanges currently do not have that verifiability or decentralisation. We have accepted centralised sequencers as a default for perpetual exchanges as a trade off for speed.

Open source, verifiable markets built on public infrastructure will be tremendously valuable as the world’s assets settle on-chain. To understand why, it helps to have a perspective on what happens when markets are walled off and centralised.

1. The London Interbank Offered Rate Scandal
   LIBOR was a benchmark rate used to determine the cost at which banks paid to borrow from one another. It was determined through collecting lending rates from individual banks who in turn submitted data without verification. Since there was an active derivatives market to trade these rates, traders at banks were incentivised to provide faulty data or collude to manipulate the benchmark and profit off the derivatives. In 2012, Barclays paid $453 million to settle the allegations. Part of the reason the scandal occurred was that centralised parties (the traders) had incentives large enough to abuse the system through a liquid derivatives market.
   
   The equivalent of a LIBOR cannot happen on Aave because the lending rates are open and verifiable. The lending rates are not set by individuals, but rather by supply and demand. Anyone can audit the numbers at any time, since there is no central trader providing these rates. The blockchain provides records that cannot be falsified.
   

2. The Nickel Squeeze of 2022
   In 2022, as the war between Ukraine and Russia raged, the nickel markets began acting strangely. Nickel futures jumped 250% between March 4 and 8 that year - going from $25k to $100k per ton on the London Metal Exchange. A large entity (named Tsingshan) had short exposure that would have resulted in it owing $15 billion to multiple large banks. The LME (London Metal Exchange) had visibility into only about 20% of the entity’s nickel short exposure, as much of it was in OTC transactions.
   
   To avoid losses from Tsingshan taking down the entire market, the LME suspended trading on March 8th, wiping out $3.9 billion in transactions. The market lost its integrity and shattered the trust people placed in it. 145 years of reputation, lost on a March morning.

Libor was about rate-setting collusion. LME’s situation was a byproduct of a lack of visibility into where the exposure lay. Both instances were fuelled by the lack of transparency in financial markets. Perpetual exchanges face a similar vulnerability to centralised sequencers, which create opacity around order fulfillment.

Verifiable infrastructure is a market necessity, not an option.

Today, regulators can audit and verify these markets. But it requires human intervention, and even when the regulator is involved, there can be slip-ups. Bernie Madoff’s scheme was known to the SEC as early as 1999. The point of introducing technology in these markets would be to avoid requiring individual regulators to verify, and instead have a steady stream of audits every block.

When market infrastructure (like the NYSE) can reverse transactions, it gives undue power to a handful of people who can hurt retail investors and hedge funds that are on the right side of trade. These exchanges are themselves at risk, as they often lack visibility on the complex OTC transactions market participants engage in. Blockchains provide mechanisms to solve all three problems.

1. They bring radical transparency through making it considerably easier to verify and vet where exposure stands.

2. The immutable nature of a blockchain ensures transaction finality. Settlement finalities are a guarantee unless the network forks.

3. They allow participants to verify and vet how orders were sequenced to ensure fairness in the system

Perpetual exchanges in crypto are not entirely open and verifiable. Many of the largest names have centralised sequencers whose order sequencing logic and history cannot be verified. As our markets turn increasingly reliant on blockchains for settlement, there will be primitives that aid in making markets open, verifiable and trustless. Part of our interest in Drift emanates from this view. Unlike many of its peers, the way it sequences orders, its codebase, and its liquidity sourcing can all be verified. Besides, within the mix of assets one can pick within Solana, Drift is mispriced for the traction it holds

Understanding Solana’s Perpetual Ecosystem

Solana accounts for less than 5% of all perpetual exchange volume. On an average day, 60% of its current activity is on Pacifica. Solana’s early foray into perpetuals did not go quite well. Mango Markets was exploited early on. The FTX crash left a nascent DeFi ecosystem bleeding. Much of the activity on Solana in the quarters that followed was in the spot market. This is where players like Jupiter, Meteora, and PUMP earned their mettle.

Hyperliquid, in contrast, focused on the perpetual meta aggressively when it launched. It listed meme asset perpetuals faster than centralised avenues did, thereby attracting a core subset of retail users. Combined with marketmakers that benefited from the exchange’s faster settlement times, this created a nascent but healthy ecosystem. It set the stage for what we know as the decentralised perpetual ecosystem. As of writing, Hyperliquid accounts for ~30% of all volume on decentralised perpetual exchanges, compared to ~5% on Solana.

Jupiter is a leading contender for perpetual products on Solana. It is hard to compete against the network effects the protocol has from its multitude of products. It has also issued JLP - an index instrument that represents BTC, SOL, and a mix of stablecoins. Jupiter’s higher TVL and take-rate make it interesting, but a bet on Jupiter is a bet on a generalist that has too much to do. We had earlier explored Jupiter in this article.

Jup has generated over $1.2 billion in fees from its perpetual product alone. The challenge is that much of the fees go back to liquidity providers on the exchange. Jup’s token model passes on fees in the form of USDC in yield back to token holders. 66.7% of fees generated on Jupiter go back to liquidity providers on JLP. Think of this as the cost of capital for TVL on Jupiter’s exchange. The product has capital density but pays a heavy premium for it. 22.2% of the fees go toward the protocol, and the remaining 11.1% is distributed to users as staking yield. So even though the perpetuals product generates a lot of fees, much of it is not captured or reflected in the token.

A different way to see it is through the lens of take rates. As seen in the chart above, Jupiter’s take rate on its perpetual exchange is 16 times what Drift charges. This is both a feature and a bug. Jupiter’s fee rate steadily climbs due to its linear pricing mechanism for its perpetual exchanges. Large orders that move markets tend to pay more. It is a good strategy as the exchange has made verifiably more than Drift in the past year. For large institutional traders & high frequency traders, the model will simply be ineffective as pricing cannot be predicted.

The parallel here is that of Google and Yahoo in the early 2000s. Yahoo branched out to multiple product lines considerably before Google did. The former focused on its core business (advertising) before expanding into email (Gmail), streaming (YouTube), and operating systems (Android). Yahoo expanded early through costly acquisitions like Geocities and Broadcast.com, but struggled to integrate them into its core business. Google focused on search and kept the product simple, as that is what the users wanted. Today, close to 80% of Google’s revenue is its core search and advertising business. Another 11% is from its cloud business. Only 9% of its revenue comes from the expanded product line.

Pacifica is the new upstart on Solana that has seen more trading volume than both Jupiter and Drift. Despite being in closed beta, the product generated over $6.7B in volume, compared to Jupiter’s $3.3B and Drift’s $ 1.4B. It accounts for nearly 60% of the total volume in Solana’s perpetual exchange ecosystem. However, there are two key concerns. Firstly, it has a centralised sequencer that is neither open nor verifiable. Secondly, the liquidity model used is that of a CLOB. If we were pursuing a native CLOB exchange, our bet on Hyperliquid would already cover that side of the market. Between these two points of friction and the fact that Pacifica does not have a liquid token, we had to explore further.

For us, the composability, verifiability, and openness of Drift is worth a premium.

Jup, Pacifica, and Drift are three entirely different products when it comes to perpetual exchanges. For JUP, the perpetual exchange is an extension of their product suite, and the fee model is optimised towards yield. For Pacifica, the focus is on being a CLOB product that is at parity with centralised exchanges. For Drift, the focus is on composability, verifiability, and openness. We go deeper into Drift because of how its liquidity model could scale, as the mix of asset types on Solana evolves from where the ecosystem is today.

Where Drift Differs

Drift is a perpetual exchange focused solely on improving infrastructure for margin trades and can iterate faster than an aggregator. The market tends to discount Jupiter for such reasons. Drift is also home to the second-largest TVL among perpetual exchanges. It holds close to 15% of all TVL committed to perpetual exchanges across all chains.

Drift has not seen meaningful competition from the emergence of other perpetual exchanges on Solana, either, until Pacifica came. Jup and Drift collectively own 95% of the TVL in perpetual exchanges. Drift’s share has stayed steady over the course of the year.

Drift trades at a $95 million FDV, compared to Jupiter’s $1.05 Billion. Core to our thesis on Drift is how its liquidity model works. Conventional exchanges (like Hyperliquid and Pacifica) follow a centrally sequenced CLOB model (CLOB stands for central limit order book). The exchange holds the power to determine how individual transactions are ordered.

In traditional markets, having the power to outrun the way orders are routed to an exchange is a multi-billion-dollar endeavour. Flash Boys by Michael Lewis famously exposed how a handful of firms had an undue edge through latency arbitrage. While CLOB-based models are necessary to compete with CEX-like features, the cost is a lack of verification and real-time settlement.

This is where Drift’s liquidity model becomes interesting.

When a user sends an order to Drift, the order routes through three forms of liquidity.

A Just-in-Time Auction

A Just-in-Time Auction Orders on Drift has a 5-second phase during which multiple market makers can competitively bid to fill an order. This is similar to the RFQ system we had written about bebop here. RFQ systems benefit consumers by enabling large orders to be competitively bid among multiple parties. It helps a market-maker, as they have the option to avoid toxic flow and reuse idle assets in external venues where they could be more productive.

In a CLOB model, consumers see the price and spread they could expect. A JIT model can absorb more in both volume and size at tighter spreads, even when the order book does not appear highly liquid. The downside of such a model is the time it takes for orders to settle. Both market-makers and traders are ultimately settling on Solana, so there will be an infrastructure bottleneck.

The other assumption required for this to work is that a multitude of market-makers want the order flow from such an auction. If competing platforms offer points or rebates alongside high amounts of volume, market-makers may have little reason to bid on Drift’s JIT auctions.

A DLOB maintained by keepers

If no market-maker bids on the order, a network of keeper bots financially incentivised by Drift can place bids. Rebates from Drift’s trading fees are passed on to keeper bots. Individual keepers are similar to brokers on the NYSE. Except that each of them can verify and maintain their own version of the orderbook permissionlessly. Everyone keeps a copy of the orders, open for anyone to verify.

A decentralised limit order book orders transactions by price, age, and size. When trigger conditions are met, the bot fills the order, depending on how it is configured. Drift - the protocol itself does not dictate how orders are sequenced. Everyone has a copy of the ledger because orders are settled on Solana.

Individual keeper bots have the ability to arbitrage against third party venues or trade among themselves to give the trader a bid. Their primary role is to keep a record of all orders. The DLOB keeper bot model is open-source. Anyone can spin up their own bot, run it, and thereby verify any of Drift’s claims. Verifiable.

A Virtual AMM

In edge scenarios, market makers can simply choose not to bid (in JIT), and keeper bots can avoid providing liquidity. This is what usually happens on days when liquidity cascades take markets down. Given that, in CLOB models, liquidity is a byproduct of humans (or APIs) providing it, traders can be left without a guaranteed bid. Drift’s approach to the problem is to provide a virtual AMM pricing model (similar to Uniswap’s) that uses user collateral and protocol balance sheet assets (liquidity provider deposits and the insurance fund).

It ensures that there is always a bid for traders, although spreads may widen and slippage increase. This has proven to be effective in scenarios where auto-deleverage kicks in and liquidates users on a CLOB-based market.

This aspect of Drift is currently underpriced by the market. Most exchanges that operate on CLOBs offer instantaneous settlement on the underlying network, but the order sequencing itself is neither verifiable nor composable. The use of perpetual exchanges cannot occur without infrastructure that is scalable and verifiable. Drift’s core architecture can deliver on these promises while incurring some sacrifices in settlement time.

As real-world assets like gold, equities, and commodities come on-chain, markets will prefer avenues that offer these benefits over recreations of the centralised structures that already exist. The ones that dominate CLOB models will be avenues that compete on liquidity, speed, and depth. All of those metrics are currently dominated by Hyperliquid. Drift sits on the other end of the spectrum, where a relative lag in an order going through is fine if the transaction sequencing itself is verifiable.

This positioning of Drift is tremendously valuable due to a few core factors that collide

1. New instruments on Solana are yet to scale. Internet capital markets in their current format are at best a meme. Startups and founders will raise capital on the network for longer-standing assets. MetaDAO is a sign of this shift. These instruments will need a perpetual exchange to enable a rich capital market. Lending needs hedging. Margin allows risk-on exposure. Spot trading allows a singular user persona
   

2. Real-world assets - like equities, foreign currencies, and tokenised real estate will increasingly come to Solana given its heavy retail-oriented user base and strong developer base. These markets will need a decentralised avenue to hedge. Composable workflows, such as taking a loan in USD and hedging against currency risk on a perpetual exchange, do not exist yet. Drift will be able to capture a reasonable share of this market.
   
   

3. Commodities - like gold, silver, and platinum are yet to find a market (of scale) on Solana. Decentralised.co’s own portfolio company (Orogold) is working on the problem. If commodities are to scale on-chain, market makers will need an avenue to hedge exposure with minimal risk on the same network. This is where Drift can play a role.

Ultimately, Solana benefits from a healthy ecosystem of traders, developers, and capital allocators that seems to be struggling to find a shared language for discussing risk. Perpetual exchanges are that language. Drift is where the language is spoken.

One way to see Drift’s potential for growth is by comparing Solana’s active user base with Hyperliquid’s, as shown in the chart below.

Assuming even 2% of Solana’s existing user base turns to Drift Protocol as users, the number of active traders on Drift will be in parity with what Hyperliquid has. When new primitives launch, they generally focus on a single application. Hyperliquid focused on trading for over a year before enabling lending. But the core user base of traders does not need lending at scale. The broader ecosystem within Hyperliquid struggles to capture the core users who came there to trade.

Drift, in contrast, benefits from Solana’s varied user base.

Solana users have the choice among NFT collections, gaming, or payments, and a meaningful share of them will find themselves on Drift. The exchange benefits from the organic, cross-category demand that a single-purpose chain may not see. This assumption, however, requires the asset types on Solana to evolve beyond where they are today, from meme assets to primitives that enable capital formation at scale.

We already see primitive versions of these trends in our portfolio companies. Asgard.fi mimics a traditional prime brokerage through routing lending positions across perpetual exchanges to generate yield. Close to 45% of their overall volume and 95% of their volume on JLP go through Drift today. Asgard could offer this infrastructure to a traditional fintech player focused on on-ramping, creating a neo-banking experience where users can earn yield on their dollars through such composable layers.

Given the network effects of existing Solana-focused service providers, we believe the surface area for such opportunities remains underexplored.

While these statements about product direction are forward-looking, it is important to note that Drift’s team is one of the few that have consistently built a product with meaningful revenue, across cycles. Q4 2025, in fact, marked an ATH for multiple exchange metrics. Compared to Q2 2024, earnings increased from a paltry $970k to $6M on the exchange. Its average daily volume doubled - from $126M to $268M, and the total volume traded on the exchange scaled from $11B to $25B.

The Revenue, Valuation Divergence

Drift remains undervalued when compared to its historical trends, too. Drift is trading at the lowest multiple (on fees) it has ever had. At its peak, Drift’s price-to-earnings ratio was 80. It generated $5.7 million in earnings in Q4 2024 and traded at an FDV of $1.85 billion. ( We annualise the quarterly figure to $23.1M for earnings) Last year, Drift generated a total of $19.8M in earnings. It trades at an annualised protocol earnings of ±4.8x. If it were to simply return to its all-time high multiple of eighty, it would trade at an FDV of $1.58B. A more conservative and realistic outcome would be a return to a multiple of twenty (based on current earnings of $19.8M), which would have the token trading at $398M, or four times its current level.

Another way to understand this divergence is by observing the difference between current volume metrics and past highs. When the market is in favour of an ecosystem, it can offer an extreme premium (such as a 40x multiple) relative to Drift’s current valuation. We model two scenarios. One where Solana’s perpetual exchange volumes surpass the $40B it sits at today. And one where Drift is able to return to capturing 60% market share within Solana’s ecosystem. In both scenarios, we use fees to measure multiples as the protocol’s retained earnings may change in the coming years.

There are three catalysts that will drive valuation up.

1. Traders who left for Pacifica to receive airdrops will return to Drift. Users routinely switch between products in anticipation of token rewards. Over time, users go where a pre-existing relationship and better liquidity exist.

2. Secondly, the nature of RWA assets available on Drift expansion will contribute to its overall valuation.

3. Lastly, Drift’s presumed 2.2 bps in our modeling is extremely conservative. As shown further above in the document, that metric is usually twice as much and closer to 4 bps on average.
   

Solana is steadily positioning itself as the layer on which all fintech primitives will emerge. There needs to be an avenue for hedging assets, for price discovery to occur, and for traders to have easy access to leverage. Drift sits at the intersection of all three needs. It will see positive tailwinds as Solana’s focus shifts from being mostly on crypto-native spot assets to a mix of commodities, indices, and equities in the coming quarter. We see the world coming on-chain, through Solana. The data below explains our reasoning for it.

Solana is undergoing its own infrastructure evolution. If Alpenglow, from Anza, comes to market, transaction finality will happen in roughly a third of the time (150ms versus 400ms currently) directly reducing the latency cost that JIT auctions currently carry. A frequently debated point of view is that Solana’s transaction throughput is the bottleneck for perpetual exchanges. Alpenglow changes that equation on finality.

Firedancer, an independent validator client built by Jump Crypto addresses throughput through parallelisation. Once live, Solana’s ecosystem will have two independent validator clients built by entirely separate teams. This matters specifically for Drift because its verifiability argument only holds if the chain it settles on is itself resilient. In the meantime, Drift’s gasless trading features have already pushed the product closer to parity with its peers on speed of execution

The World will Trade Onchain

Solana’s RWA base has grown from a little over $10 million in January of 2024 to over $1.6 billion in the span of two years. It is far smaller than Ethereum’s $14.7 billion, which accounts for ~59% of all RWA assets represented on-chain.

According to RWA.xyz,

1. Solana has the second-highest holder base of RWA assets on-chain at 283k, compared to 160k on Ethereum

2. It was home to 60k active addresses engaging with RWAs on-chain in January of 2026 alone.

Solana has been where retail participants already are. It is where they come to hold meme assets, send dollars cheaply, and stay up-to-date on culture. The fact that these users currently generate $120k each day on Phantom through builder codes is a sign of how big the market for retail participants trading perpetuals could be. There is a segment of the world that needs access to commodities, equities, and foreign currencies without going through local brokers. HIP-3 serves the crypto-native crowd quite well today.

But when you consider tooling such as wallets, off-ramps, and private key management solutions, along with the mix of spot assets, Solana is positioned as a better ecosystem for developers to build on. We are already witnessing an uptick of RWA assets on Solana.

Currently, xStocks alone does $600 million in on-chain volume in spot markets each month. HIP-3 native markets are expected to reach $100 billion in the coming months. Presume half of that would trend to Solana and be worth $50 billion each month. Given Drift’s market positioning, it could command half of that volume - and capture $25 billion. Assuming a fee rate of 2.2 bps on that volume, we are looking at a fee uptick of $5.5 million per month. The number is 5.34 times what Drift reported in earnings in January, at $1.03M. Annualised, that amount will be $66 million. Considering a 5x PF, we are looking at a $ 330 M valuation.

Drift can capture half of the RWA segment volume as it has demonstrated the ability to capture 60% of the market on Solana in the past, and its architecture is meaningfully better for institutions that care about order sequencing, settlement times, and verifiability.

In 2025, Drift was early to collaborate with Securitize to bring a tokenised version of Apollo’s Credit fund on-chain. They have proven their ability to structure, on-board, and create MVPs with the largest asset issuers to extend asset representations on-chain. As the demand for RWAs increases, Drift benefits from its storied past and strategic positioning within Solana.

Assets on Solana usually trade at deep discounts to their fees due to the seasonality of revenue sources or being relatively early. Drift, by comparison, commands a premium due to its steady, consistent governance structure and the fact that the team has silently built a cash-flow machine that sustains itself through market cycles across all market environments. Everyone needs a perpetual exchange regardless of whether tokens go up or down. Drift has consistently been the best avenue for traders on Solana. It has the opportunity to simultaneously expand its volume base while commanding a premium in the Solana ecosystem.

If it is to establish itself as the home of all DeFi on Solana, we expect this premium to grow substantially. Solana has yet to respond to Hyperliquid.

What Could Break

For Decentralised.co - the bet is not about going against Hyperliquid. It is a long on Solana’s ecosystem through a player that is deeply discounted for the value it generates and the premium its team should command. There are several assumptions that can break here.

1. As it stands, Drift’s revenue is being used to sustain the team. This is considerably better than selling tokens to fund operations, but if other perpetual exchanges continue to command volume and Drift struggles to catch up, it may find itself having next to no capital to reinvest in the community or product through incentives.
   

2. We also presume that price-to-earnings ratios are good indicators of a network’s value, but over the past four years, owning the token has not meant much in terms of how earnings flow back to users. The assumption here is that investing earnings in the team’s operation is currently the best use of them, but that assumption can falter.

3. Hyperliquid trades at a 29 PE because much of its earnings flow back into the token through buybacks. Currently, Drift’s model does not have a buyback program. The markets may continue to punish this through valuation discounts if there is no confidence that the product is better off leaving the fees unused.
   

4. Lastly, Jupiter, given its aggregator model, can compete with Drift on several fronts. Ultimately, the choice between Jupiter and Drift boils down to a specialist platform vs an aggregator. Our assumption is that a specialist will beat an aggregator during this phase of the market, where a sectoral leader is not yet established in Solana. Jup’s new chain and ability to acquire players within Solana ecosystem are meaningful threats to Drift.

One challenge we see is with tying the token’s revenue to its valuation. This will require working with the team on governance and on the long-term relationship between token holders and the business’s core fundamentals. As it stands, the token could benefit from more proactive governance and engagement from its holders.

For investors looking to build exposure to perpetual exchanges, Hyperliquid could be a better choice, given that the assistance fund on Hyperliquid routinely buys back HYPE tokens from the market. One of the things we would like to work on in the coming months is to propose a framework for Drift’s token to accrue more value from the traction the protocol generates.

Drift’s team has shown exceptional commitment towards building open, verifiable markets. They have been at the cutting edge of delivering a better user experience (via solutions like gasless trading) without compromising decentralization. Although markets do not currently talk about the need for verifiability or trustlessness, we believe it is only a matter of time before these elements become important.

In the past four years, Drift has shown commitment towards that mission. Our investment is a bet that as Solana’s capital markets mature, venues that prioritise verifiability, composability, and openness will command a premium.

We are ultimately long open, verifiable, composable markets. Drift is a position that expresses that conviction.

Verifying markets,
Joel John

Today’s piece might scream existential crisis, but hope is what it delivers. With VCs leaving the industry and founders wondering whether AI is the play to make, now is a good time to explore why we are even here. We broke down the numbers and traced where revenue flows. Today is part one of a series of articles that speak to the crux of why crypto matters, where it stands, and how it evolves from here.

Decentralised.co is actively investing in the liquid markets. The research that shapes pieces like today’s informs how we allocate. If you are an investor or a founder looking to explore how fundamentals tie back to crypto assets, join us at Telegram. Or drop us an email at venture@decentralised.co.

Let’s dig in..

The fear and greed index for crypto is at the lowest it’s ever been. At the same time, the industry has never been more profitable. Since 2018, DeFiLlama has tracked $74.8B in fees generated by crypto-native protocols. Nearly half - $31.4B was generated in the eighteen months between January 2024 and June 2025.

How is it that an industry can have its best series of quarters in eight years and still find itself in fear?

Entropy Protocol. Milkyway Protocol. Nifty Gateway. Rodeo. Forgotten Runiverse. Slingshot. Polynomial. Zerolend. Grix Finance. Parsec Finance. Angle Protocol. Step Finance. These are twelve names that shut down over the past two months. Products that have sustained for years and were built by passionate founders we respect OKX, Mantra, Polygon Labs, Gemini and Binance have each had layoffs, too.

Conferences have fewer people visiting. VCs are pivoting to AI. Developers are flocking to AI. The doom and gloom are real. If you are in crypto, pivot to AI - has become the prevailing sentiment.

But should you?

We have been thinking about this question for the past few weeks. When a technology emerges, markets initially assign it a premium due to its novelty and the scale of its ambition. Close to 6% of the UK’s GDP was invested in rail stocks in the 1800s. Hyperscaler capital expenditures will represent 2% of US GDP in 2026. But as reality hits, the technology trends towards more reasonable valuations. What matters is whether an industry can prove itself useful in the return to sanity.

In today’s issue, I break down how crypto revenue has historically evolved, the stickiness of the money generated, and the nature of moats in our industry.

Studying The Ledger

Crypto-native businesses have generated revenue since the industry’s inception. Exchanges like Bitmex, Binance and Coinbase were massively profitable endeavours. They were centralised, owned by a few individuals, and their revenue was not public. DeFi native primitives like exchanges (Uniswap) and lending platforms (Aave) changed the mood. You could verify how much the protocol made each day.

Tokens were expected to trade at valuations that reflected the economic activity these primitives enabled.

As late as 2022, decentralised exchanges accounted for 28.4% of revenue generated. Together, they generated $2.27B that year. Lending told a similar story and was highly concentrated. Aave and Compound accounted for 82% of all lending fees. There were category leaders, but there was also hope in the long-tail of protocols evolving to capture market share.

The technology itself was novel enough to command a high valuation.

Crypto’s consumer expansion closely followed. NFTs represented a hopeful vision of culture being priced on-chain. Celebrities that the average person knew were changing display pictures on Twitter. People presumed this would translate to massive adoption. OpenSea generated $1.55B or 71.7% of all NFT market revenue. In hindsight, the marketplace’s $13 billion valuation looks less absurd. They could have evolved to be a sustained monopoly.

Fate and markets had other plans. In 2025, NFTs accounted for less than 1% of revenue. We had our Beanie Baby moment, but it left us with no physical memorabilia. Decentralised exchanges, in contrast, grew but struggled to capture any valuation gains. Last year, decentralised exchanges generated $5.03B in fees. Lending platforms generated $1.65B in fees. Combined, these two segments account for 22.9% of overall fees, down from 33.1% in 2022.

Their economic activity represents a smaller share of a larger pie, but their valuations have dropped drastically.

What grew then?
How did crypto’s native business models evolve in the years since 2022?

The chart below offers clues.

In January 2026, stablecoin issuers Tether and Circle accounted for 34.3% of all fees generated. Put differently, $0.34 of every dollar earned in the industry goes to two firms. Their revenue has doubled from $4.95B in January 2023 to $9.89B in 2025, driven almost entirely by T-bills. These are startup metrics for bank-scale financial products. Tether generates almost three times as much revenue as Circle.

Their rise came down to two forces.

1. The first was demand. The Global South always needed instruments to hedge against local inflation and to move money freely. Dollars, even digital ones, fill that gap in a way local currencies cannot. Capital flight is a necessity. Not a feature.
   

2. The second was the cost structure. Blockchains absorb the operational aspects of running a stablecoin business. Unlike traditional banks or fintechs, Tether and Circle do not need to hire in proportion to the amount of stablecoins issued on-chain. The marginal cost of issuing the next $1 billion on-chain and moving the next $100 billion between addresses is next to zero.

The two forces tangoed. One was pulled from the demand side as citizens voted with their money. The other flattened the cost curve. Together, they made stablecoin issuance one of the most capital-efficient businesses in financial history.

Stablecoin businesses require moats in liquidity, compliance, and the Lindy effect. The number of issuers that have survived multiple cycles is incredibly small. Between Tether and Circle, close to 99% of all stablecoin issuance revenue is captured. Why is that the case? Both assets benefited from being early. The network effects of multiple exchanges plugging in legitimised them in a way technology could not.

Tether had initially launched on Omni as a side-chain. It was slow and clunky, but it was also accessible in the touchpoints OTC desks and exchanges interacted with. That was a distribution moat, not a technology one. A moat that crypto-native founders often struggle to replicate with code alone.

Stablecoins benefited from the Lindy effect.
A new category would soon benefit from distribution moats.

We teased the idea that crypto is a transaction economy in two of our pieces. One was Money Moves, and the other was written last year in Everything Is a Market. What we did not account for at the time was the pace at which transactional products built around Telegram trading bots and trading interfaces were growing.

Combined, in Jan of 2025, these two sectors alone accounted for $575M in fees. It makes sense when you consider what consumers want. Meme trading and perpetual exchanges allow users to make quick profits. They pay hefty fees in pursuit of quick returns. The category rose from 1% of overall revenue to just over 15% between 2022 and 2025.

Products like TryFomo and Moonshot generated millions of dollars in revenue by focusing on the end user. The product was not technically sophisticated. On the contrary, it was the ability to aggregate crypto-native primitives and bundle them to create better user experiences. Developers no longer had to incentivise liquidity or bother about wallet management thanks to tooling like Privy coming of age.

The primitives that we were all excited by in 2022 were now mature. Applications (such as BullX and Photon) were built on top of it. The sector alone created some $1.93B in fees between January 2024 and February 2026.

Meme assets had one fatal flaw. They were thin applications with great amounts of seasonality attached to them. Feels like Déjà vu? It is because NFTs and Web3 gaming had similar spurts of growth and eventual collapse. This seasonality is both a bug and a feature in our industry. We will revisit this theme shortly. But for now, let’s try to map out where revenue went instead.

Perpetual exchanges (and later prediction markets) represented new avenues with longevity. PumpFun democratised asset issuance through meme coins, but the game was not fair.

Eventually, markets wised up to the fact that meme coins die. Dreams of being a millionaire through buying a token named ShibaInuYouShouldShareThisNewsletter coin died with them. People did not want to manage a portfolio of random tokens. They wanted risk. Perpetual exchanges offered that.

Share

You could trade Bitcoin, Solana or Ethereum with extreme leverage. Market-makers and traders that needed an alternative to centralised avenues flocked in. The core product in this category was liquidity. Hyperliquid was the dominant player because its orderbook depth was comparable to that of a centralised exchange. Without that parity, users had no reason to move over. Hyperliquid and Jupiter accounted for the bulk of the fees in the category over the past three years.

Perpetual exchanges and trading platforms ripped the band-aid off crypto. They made it apparent that capturing small sums of fees off high-velocity transactions was where the money had to be made. Meme trading platforms and perpetual exchanges were dopamine machines that packaged and sold risk.

One of them would mature to be a core financial primitive. One that the world would use to trade commodities, equities and digital assets even on a weekend. Blockchain-native apps replicated what Robinhood and Binance had long offered: avenues for risk.

Starving a Fat Protocol

Notice that I haven’t mentioned protocols so far? The base layer where all the magical internet money movements were recorded? That is because they have an entirely different (but pertinent) story to say. They are casualties of the novelty premium, which is trending towards a discount. Let me explain

In January 2023, Optimism was trading at PF (price-to-fee) multiple of 465. Solana was at 706. Arbitrum and BNB were around 206. Today, Solana is at 138, Arbitrum is at 62, and OP is at 37. Polygon trades closer to a fintech company at 20. Tron sustains the stablecoin ecosystem and sits at a multiple of 10.2. Taken individually, Optimism, Solana, Arbitrum, and Polygon have enabled more sophisticated products in the years since. Each of them has more users, better liquidity and a more complex suite of financial applications built on top of them.

The discount in their price-to-fee multiples tells us about how the market perceives them.

Historically, L1s and L2s have traded at extreme premiums compared to standalone infrastructure or ventures. That premium, when invested well, could have created new economies. It could have funded developers to build real applications that mattered to real people outside our industry. Instead, the open-source nature of the products and the ease with which things could be tokenised meant we had fifty replicas of the same thing across thirty networks, with broken composability.

That was okay, because we had bridges. And cross-chain messaging. And countless other mechanisms to move money. All of which were now being valued at increasingly lower levels.

Consider the fate of DeFi primitives. Investor optionality and the lack of novelty butchered valuations, even when these primitives drove more economic activity. These are fragmented markets where investors have a multitude of choices to bet on. The novelty of being “decentralised” or on a blockchain has long since worn off. Kamino, Euler, Fluid, Meteora and PumpSwap have entered the chat - all at lower price-to-fee ratios than what protocols commanded in 2022. As the chart below from TokenTerminal shows, price-to-fee multiples of decentralised exchanges collapsed between 2023 and 2025. Some now trade at a multiple of 1.

In other words, the market values them below the fees they will generate over the coming year. A strange paradox is on display. Although the underlying protocols, be they DeFi primitives or the L1 itself, have trended lower in valuation, the apps built on them have generated more revenue over shorter periods.

The number of teams doing over a million in revenue every quarter has steadily risen since the early 2020s. The figure is over one hundred. Protocols that took twenty-four months to reach $10 million in annual revenue in 2020 were considered fast. By 2024, protocols were reaching that milestone in approximately six months. Pump.Fun, launched in early 2024, reached $10 million in revenue in roughly two months, the fastest recorded trajectory.

This acceleration reflects both the maturation of the underlying infrastructure (faster chains, cheaper transactions) and the expanding pool of on-chain capital looking for yield and entertainment. If you’re a developer or a founder, consider the facts:

• There are nearly 900 protocols generating revenue in crypto today.

• Each of them competes for a smaller share of median revenue, but the broad arc shows more teams generating more revenue. For context, revenue-generating protocols grew almost eight times, from 116 to 889.

• Median revenue has declined to $13k every month

Blockchain native businesses have three forms of moat. Each of which is evident when we study how revenue is made.

1. First Mover Advantages: The network effects that Tether and Circle gain from being early will be hard to replicate. They have been around through multiple cycles and are established duopolies, despite the emergence of new players. As it stands, these businesses are non-tokenised and hyperfinancial. Tether is a centralised entity whose revenue mostly comes from T-Bills.
   

2. Liquidity Moats: Aave has sustained liquidity depth across cycles in an industry where capital has historically been mercenary. Hyperliquid has replicated it, but it is too early to say. These protocols have an incentive to return capital to liquidity providers and tweak tokens towards governance.

3. Distribution Moats: Seasonal applications, such as meme-coin trading platforms, rely on capital velocity and consumer demand. Web3 games and NFTs stick here. AI-fuelled productivity would mean small, lean teams can now ship consumer-oriented products faster. Where does edge come from? It would boil down to onboarding and retaining the most users when markets are hot.

Products built on distribution moats can be tremendously valuable, but they are outliers and not the norm. A startup was traditionally valuable because its learnings could be replicated. Y Combinator works partly because of the lindy effects of the ideas that prevailed. Crypto evolves at a pace that is too fast for such Lindy effect learnings to be instilled. This is partly why we do not see founders replicate their success in consumer products as a category. The seasonality that helped scale a venture in the first place may not replicate itself.

That is not to suggest founders should not pursue the opportunity. Segments such as prediction markets or data providers for agentic-economy products may generate a great deal of cash flow in the short run. It is just pertinent to understand that they are high-velocity, short-term games that may not last. The trap for such product categories is in needlessly raising venture capital or being stuck with a token issued long after the meta that enabled the product in the first place dies.

What makes tokenised businesses valuable then? Are their valuations even justified?
The data offers some clues.

Questioning Governance

In 1999, many tech companies traded at a price-to-sales (P/S) ratio of 10x-20x. Akamai, a content delivery network, traded at 7,434 times sales. By 2004, Akamai had fallen to 8x. Many companies went from P/S of 30x-50x to less than 10x. The dot-com bubble burst evaporated trillions of dollars in speculative value. Yet many of the companies survived because the underlying businesses were real. Amazon fell 94% from its dot-com peak and went on to become one of the most valuable companies in history.

Crypto is going through the same compression, faster. In 2020, when DeFi was still an experiment, crypto generated just about $21 million in total annual revenue. The aggregate P/S across tracked protocols sat at 40,400x. The market was all about the future: “What could crypto be?” By 2021, as DeFi summer turned protocol revenue into real numbers, P/S collapsed to 338x. Today, with $18 billion in annualised revenue, it sits at ~170x. The compression from 40400x to 170x took five years.

However, there is a catch. When Visa trades at 18x, shareholders receive dividends and buybacks. They have a legal claim on the company’s earnings and a seat at the governance table enforced by securities law. When Aave trades at 4x, token holders have governance rights but, until recently, no direct economic claim on revenue. Hyperliquid uses its assistance fund for buybacks, making HYPE holders the closest thing to equity holders in DeFi. Aave approved a $50 million annual buyback programme in 2025.

These are meaningful moves, but they are exceptions. Across the broader market, most protocols lack a mechanism to return value to token holders. The multiples look cheap, but the rights attached to those multiples are thinner than anything in traditional markets. What makes these multiples possible is an industry that generates revenue at a scale and efficiency unlike anything in traditional business.

The protocols compressing crypto’s P/S ratio are not large organisations with thousands of employees. They are small teams running global financial infrastructure with near-zero marginal costs and no physical footprint. How thin can those costs get? And how much will holders trust these teams with doing the right thing with protocol revenue?

Segregate the market by sector to get a much better picture. Aave, the largest lending protocol in DeFi, trades at a P/S ratio of roughly 4x. Hyperliquid, which controls ~80% of the decentralised perpetual futures market, trades at roughly 7x. These are not bubble multiples. Arguably, they are lower than their closest traditional comparables. Coinbase, the only major publicly listed crypto exchange, trades at roughly 9x revenue. CME Group, the world’s largest derivatives exchange, trades at ~16x. Visa, the gold standard of payments infrastructure, trades at ~15x.

Will Clemente mentioned on our podcast that crypto is the purest form of capitalism. In no other industry do successful businesses achieve profit-per-employee numbers that approach Tether’s estimated $100 million. To put this in perspective, Nvidia generates $5.2 million in revenue per employee, Apple generates $2.4 million, and Google, $2 million. Tether, with 125 employees and roughly $12.5 billion in annual revenue, operates at a scale that suggests the highest profit per employee in corporate history.

Despite the crazy 170x P/S number, the market is not irrational about the protocols that actually generate revenue. It prices them at or below traditional financial infrastructure.

Which leads me to the next question - what even is the use of tokens? In many categories, tokens are powerful instruments for coordinating capital towards a shared vision. Crypto is in this phase where entrenched duopolies are the norm. Traditionally, a founder had to take debt (against equity) or raise money to infuse capital for financial products. Hyperliquid, Uniswap, Jupiter, and Blur are proof that, with token incentives, individuals will allocate capital to new products. The same individuals could contribute substantially when tokens have governance rights attached. In this regard, tokens may evolve to serve two functions.

1. Firstly, to coordinate capital and resources from the right kind of individuals

2. Secondly, to empower them to govern a protocol.

Tokens on their own are no longer valuable. Even equities are tokenised now. These instruments will need to have claims to economic activity and the ability to guide governance. Many Layer 1 and Layer 2 tokens struggle to deliver on either. Teams and VCs often hold the bulk of the token, leaving holders in disarray. It leaves no reason for the marginal person to care about newly listed digital assets.

Today, these experiments are split. MetaDAO allows holders to receive a full refund of their investments if teams make false representations. There are no large protocols on it yet. Crypto’s core reckoning is that tokens have traditionally given holders few rights. Protocols are now trying to answer the age-old question of why anyone should hold these instruments in the first place. In a future issue, we will explore the correlation between holder rights and valuations.

The Fork In The Road

Capital markets have become more intertwined over the past two decades. Much of it can be attributed to technological improvements. We can trade commodities, foreign indices, digital assets and even compute (GPUs) in the near future. Blockchains have made it possible to trade these markets globally, at all times. The fact that Nasdaq and NYSE are now trending towards 24/7 markets is an instance of a technology changing the zeitgeist.

We live in a hyperfinancial world, where ironically, news of war has us scrambling to figure out which prediction markets are the best to bet on.

For founders, that means rethinking what they build and how they build it. If the data here makes anything apparent, it is that all blockchain products eventually monetise through two core maxims.

1. Through taking a small cut of high-frequency transactions or

2. Taking a large cut of transactions where verifiability and trust assumptions matter

The edge is either in velocity or verifiable transparency.

Profit motives are the purest incentives for participants in the capital market. There is a core belief that eventually, markets trend towards extreme efficiency. We see versions of this with categorical leaders. That chart where we saw 70% of multiple segments being owned by two key players? That is a brutal aspect of life and markets we all grapple with. What this means for founders is that the capital that once went towards their token is now redirected to assets with higher volatility or higher returns on capital.

Long-term capital exists, and may even pay a premium, but only if it values the underlying business. Investors in Google and Amazon did not have to scramble for the exits because the underlying businesses were valuable.

In an age where the value of software itself is under question, blockchain native apps will have to find new ways to be valuable. We can restructure tokens. Perhaps, we could even make start-up equity tradable on-chain. But this is not just a token problem. It is a business model problem, too. The vast majority of long-tail blockchain applications - like Web3 social, identity and gaming products have struggled to see scale or meaningfully differentiate from their traditional counterparts. It is not that those experiments were not valuable. It is that we struggled to monetise them effectively.

The infrastructure age of crypto is behind us. What lies ahead is a period when it blurs with the Internet. Nobody talks about an “online” business anymore; you simply exist on the Internet. Nobody is a “mobile app developer” anymore; you are a developer.

Long live the age of the blockchain enthusiast. We are simply ledger maximalists wondering what the best use of these ledgers is.

Craving a good shawarma,
Joel John

P.S - I wrote this piece, part monitoring drone hits on the town I call home - Dubai. There is a silent resilience in continuing, even when it may seem as if the world is coming to an end. A rebellion in its own right. Many founders in crypto are displaying the same resilience. If you are still here and building, we’d love to chat and see how our data and learnings could be put to use in building scalable protocols.

Research here is guided by data from DefiLlama, TokenTerminal, Codex, Santiment and a mix of traditional data vendors.

Base just announced it's moving away from Optimism's OP Stack to build its own architecture. OP is down almost 25% in a day after the largest source of revenue for Optimism chose not to pay them anymore.

We didn’t know about this when we started writing the article, but we made the case for further fragmentation in crypto. Fragmentation at the level of chains, token standards, and liquidity. Over the past few weeks, we have been working with the LI.FI team to make the case that fragmentation in crypto is not a temporary phase. Base walking away just proves the point.

If you are building across chains, trying to bring all of crypto together, write to us at venture@decentralised.co.

TLDR:

Complex systems have a tendency to fragment before they consolidate. Telephones got automatic switches. Banks got clearinghouses. The internet got TCP/IP, then search engines. Crypto followed a similar path. The industry expanded into a landscape of L1s, L2s, and app chains; bridges and DEXs now connect them. But the problem is that delivering an outcome still depends on multiple independent services behaving correctly in sequence.

LI.FI is an orchestration infrastructure for this problem. It distils cross-chain complexity into composable building blocks that developers can combine to build the flows their products require.

Three shifts make this timely.

1. Fragmentation is compounding

2. Regulation is driving more fragmentation.

3. Execution is becoming autonomous.

In a world of increasing asset and liquidity fragmentation, the winners will be systems that can scale with the complexity of orchestrating the steps needed to fulfil users’ needs.

Almon Brown Strowger was a mortician in Kansas City, Missouri, USA. His legend goes like this. Strowger was losing business to his competitor. He was convinced that his competitor was getting all his orders thanks to his wife, who worked in a telephone switching company. In 1891, Strowger built a device that would route calls without human intervention.

Whether the conspiracy was real or not, Strowger’s response changed telecommunications. A caller would express their intent (the number they wanted to reach), and the switch would figure out the physical path through the network.

Before automatic switching, every call was a negotiation with the phone operator. You waited to talk to the other person because the operator physically connected the wires. As networks grew, operators became bottlenecks. The complexity of possible connections outpaced human coordination capacity.

The switch absorbed routing into the network itself. Instead of expressing an intent to a human intermediary, the caller dialled, and the system connected. This provided a much better user experience.

Lasting complex systems tend to follow this path. When they grow beyond what individuals can coordinate on their own, coordination shifts to systems designed to manage it at scale.

We have seen this pattern repeated across communication, banking, and modern food delivery apps. The Internet itself was not immune to the problem of scaling. We are also seeing this play out with blockchains. Blockchains are about making distributed computers independently agree on a version of the truth. The fewer things they have to agree on, the better. This is why blockchains are closed worlds. Something is needed to coordinate how value and information move across these closed systems. Bridges were the first step towards this coordination.

This article examines how increasing fragmentation creates orchestration bottlenecks, how LI.FI is improving the user experience across several blockchains and is shaping the interoperability (interop) space as more and more global finance shifts onto blockchain rails.

Multiple Networks to One Internet

The Internet grew out of many small, purpose-built networks run by universities, research labs, and government programmes, each solving a local communication problem rather than trying to coordinate globally.

Until the early 1980s, computers were connected through shared institutional systems and research networks where access was limited, connectivity was intermittent, and communication depended on telephone infrastructure. These networks worked, but they were expensive and not designed to interoperate by default. As the number of networks grew, coordinating between them became more difficult.

CompuServe, AOL, and Prodigy each operated its own network. Users subscribed to one service, logged into that service’s servers, and interacted only with other users and content inside that network. Communication across services was not supported.

From the user’s perspective, this meant that “being online” did not mean accessing a global network. It meant choosing which system to subscribe to. Content did not flow freely between services. Messages sent on one platform were invisible to users on another. There was no shared discovery layer, no shared messaging standard, and no shared economic rail.

Crucially, this problem could not be solved by any single platform growing larger. Even as AOL scaled rapidly, it remained incompatible with other services. Growth improved coordination within a network, but did nothing to reduce coordination costs between networks.

The breakthrough came by standardising how data moved between machines rather than trying to standardise user experience or consolidate platforms. This led to the development and adoption of TCP/IP, a set of transport standards that defined how information could be addressed, transmitted, and reassembled across heterogeneous networks.

It allowed networks to retain their internal structure while conforming to a shared set of transport standards. Systems no longer need custom, bilateral integrations to exchange data. As long as they followed the same protocol rules, packets of data could move reliably between otherwise incompatible networks.

When communication between networks was standardised, the system could finally scale beyond bespoke integrations. Standardised transport created a shared substrate on which independent services could be built and reached reliably, even though those services were not coordinated with each other.

What was the impact of this? If there are three separate networks, you need at least three custom integrations to enable them to communicate with each other. For four networks, you need six custom integrations. Ten integrations for five networks, and so on. Statistics nerds know that it’s ^NC₂ custom integrations for N networks or systems. So if 100 different businesses wanted to exchange data, they would have needed 9,900 custom integrations. TCP/IP eliminated that need with one shared standard.

However, this did not make the Internet easy to use by default. As more organisations began publishing information and building services on top of shared protocols, the number of reachable destinations grew faster than any individual could reasonably keep track of.

The problem was no longer whether systems could talk to each other, but how users decided where to go and what to interact with. Information was technically accessible, but discovery became expensive in human terms. Knowing which site mattered, which service was trustworthy, or where value actually lived required effort that did not scale with the network’s size.

Search engines, app stores, marketplaces, and media platforms emerged to solve this by making accessing the web easier. They indexed a rapidly growing universe of independent services, applied ranking and filtering rules, and routed users towards a small set of relevant options at any given moment. So, instead of picking from millions of unrelated options, users could choose from ten highly probable options.

Search engines did nothing to unify the underlying system. Websites and services remained separate, owned by different operators. What changed was how users discovered them. Instead of users having to reason about destinations, protocols, or addresses, these platforms absorbed that complexity. They handled discovery and access at the point of interaction, allowing users to focus on what they wanted to do, from finding information to consuming content, or completing a task, without having to think about how the underlying system worked.

Crypto has gone through a similar transition. What started as a single shared execution environment has spread across many independent chains, each with its own liquidity, rules, and constraints. The orchestration problems that shaped the early internet have returned, at the level of making actions actually complete across different execution environments.

How did Crypto Get Here?

From 2015 to 2020, most on-chain activity occurred on Ethereum. Smart contracts, liquidity, and users shared a single execution environment. Assets were native, and transactions entered the same mempool and were ultimately ordered by the same set of validators. Because everything lived on the same chain, composability was the default.

During the DeFi Summer of 2020, routine actions such as token approvals, swaps, and liquidity provisioning frequently cost hundreds of dollars.

Any burst of activity delays unrelated transactions across the network.

Alternative Layer 1s and rollups both emerged as responses to Ethereum’s execution constraints. L1s reduced fees by adopting different design trade-offs, while rollups improved throughput by batching transactions and settling back to Ethereum. New chains offered aggressive subsidies to bootstrap activity.

Users were no longer tied to one chain, so applications had no reason to stay on one either. Uniswap launched on Polygon, Arbitrum, and Optimism with largely identical core logic. Aave deployed across Ethereum, Polygon, Avalanche, and later other networks. The same applications now existed in parallel across different execution environments.

As assets moved across chains, the same token began to exist in multiple places at once. Stablecoins, ETH, and governance tokens were bridged into separate liquidity pools on each network. Liquidity that had once aggregated naturally on a single chain now sat in parallel pools with different depths and conditions.

This made prices less uniform. A token swap on Ethereum might clear at a different price than the same swap on Polygon or Avalanche, simply because liquidity was thinner or arbitrage was slower. During periods of volatility, users could see materially different outcomes for the same trade depending on where it was executed.

This fragmentation also extended beyond liquidity into the state. User balances, approvals, LP positions, and borrowing relationships were now distributed across chains. Actions that were previously atomic, such as moving collateral or rebalancing positions, became multi-step workflows involving bridges, with long wait times and repeated approvals.

Bridges became the critical coordination layer. Unlike single-chain execution, bridges operate across separate security domains, relying on validator sets, multisigs, or other off-chain actors to authorise transfers. Compromising even part of this verification layer is sufficient to break the bridge, even if both chains remain secure.

This is the point at which crypto’s interoperability stack began to stratify. What had been a single problem of “how do I move assets?” split into distinct layers.

1. transport (how messages and proofs move between chains),

2. standards (how intents and cross-chain calls are expressed),

3. execution (how orders are fulfilled), and a

4. applications (how users interact with all of it).

Each layer would evolve at different speeds and face different competitive dynamics.

At a high level, bridges enable assets and actions to be used across otherwise independent execution environments. Over time, two dominant approaches have emerged, reflecting different answers to the same question: how much of the cross-chain complexity should be pushed onto users, and how much should be absorbed by the system.

On the one hand, interoperability has converged around standardised token representations and mint-and-burn models, in which assets move across chains via shared token standards and messaging layers. These systems aim to make assets broadly portable and composable across many environments, but they still require applications and users to manage execution details across chains.

On the other hand, intent-based approaches have grown in prominence, where users specify outcomes and the system orchestrates how those outcomes are achieved across chains. Instead of exposing each intermediate step, these systems shift more responsibility to the infrastructure layer.

The aim here is not to delve into bridge taxonomy, but to understand their role in the orchestration layer. Their design choices shape how failures propagate, how liquidity is accessed, and how much complexity applications must absorb.

As execution spread across Layer 1s, rollups, and Ethereum itself, fragmentation began to compound. Each environment had introduced its own sequencer, block cadence, fee market, and operational risks. Gas mechanics added another layer of friction, as users frequently arrived on destination chains without the correct gas token to complete execution. Partial solutions such as gas sponsorships existed, but they were very inconsistent and application-specific.

Teams took on this burden because blockchains were no longer just execution environments. They had become distinct markets. The chart below shows trading volume in different ecosystems. If you exclude Ethereum, there’s no clear winner. Applications could not afford to ignore other chains. Even if it were for a brief period, if your application were on a chain that attracted activity for any reason, you would stand to gain users.

Applications had a clear economic incentive to expand. Remaining on one chain meant leaving users, volume, and liquidity on the table as activity migrated elsewhere. Expanding to additional chains allowed applications to meet demand where it already existed, rather than trying to pull users back into a single environment. Put simply, if you stayed on just one chain, you left money on the table. Going multi-chain had become the table stakes.

That expansion came with significant operational overhead, but it was often the only way to maintain growth, relevance, and competitiveness in a system where usage was no longer concentrated in one place.

For developers, multi-chain execution dramatically increased operational complexity. Applications could no longer be deployed once and operated globally. Every network required its own contract instance, RPC providers, and indexers, and each introduced its own failure modes.

Moreover, this was not a one-time cost. Engineering teams repeatedly described multi-chain expansion as running the same application under different operational assumptions, requiring them to handle upgrades, monitoring, and incident response independently for each chain.

Before 1773, bank clerks physically walked across London to settle debts with every other bank. Every morning, dozens of clerks trudged through mud and cobblestones, carrying paper and coin, visiting every counterparty individually. As the number of banks grew, bilateral settlement became untenable because its complexity scaled quadratically.

The solution emerged in a coffeehouse on Lombard Street. Clerks started meeting at the Five Bells tavern at day’s end. Instead of settling bilaterally, they netted their balances. Dozens of physical journeys collapsed into one meeting. The clearing house separated what banks owed from how they settled.

Crypto today is in a similar spot, but the shape of the problem has changed. New blockchains no longer connect through simple, pair-wise native bridges. Instead, they launch already wired into the ecosystem via third-party liquidity bridges, interop token standards, and shared execution infrastructure spanning many chains. Connectivity is no longer the bottleneck.

The difficulty lies in composing actions across what is already connected. Capital can move, assets can be represented across chains, and liquidity can be reached from many venues, but every user action still depends on multiple independent systems working together. The complexity adds up in the same way it did on the streets of 1770s London.

Luckily, history offers a direction towards a solution. What is needed is not more bridges, but an orchestration layer that converges routes. Something akin to a Lombard Street for on-chain settlement, a layer that separates what users want from how it is executed. This is the gap that systems like LI.FI fill by providing a single integration surface where multiple execution providers (like liquidity pools, bridges, DEXs, fast paths, and intent-based systems) come together to fulfil actions across any chain.

The Aggregation Trap

When execution fragmented across chains, the most immediate problem was navigation. Knowing where liquidity lived, which bridge supported a given asset, or which DEX offered the best price required manually checking multiple interfaces.

Aggregators addressed this by building backend infrastructure that indexed routes across bridges and DEXs and exposed them through a single integration surface. Instead of every application needing to maintain its own view of the ecosystem, they could rely on a shared routing layer to discover viable paths.

The pain point shifted from discoverability to coordination.

At that stage, aggregation seemed the best approach for orchestration and abstraction. It aligned well with how the system had evolved. Liquidity was already fragmented. Execution environments were already independent. Aggregators only needed to observe them and present options. The most obvious friction was informational, and aggregators solved that cleanly.

However, aggregation did not improve the user experience to the desired extent. Presenting information in a single interface did not mean that all user actions could be handled with a single click. In addition to worrying about execution costs, users also had to manage gas tokens across multiple chains. They had to undergo the ordeal of acquiring gas tokens on the destination chain to complete any cross-chain action.

Let’s say you had ETH and wanted to deploy USDC in some lending pool on Avalanche. Aggregators would swap your ETH to USDC, bridge to Avalanche, but could not deposit into the lending pool because you didn’t have AVAX (the gas token) on the destination chain.

Apps like Jumper solved the gas problem. Many routes now include refuelling or gasless execution. But this only fixes one visible dependency. Cross-chain journeys still lack end-to-end atomicity. Even with gas handled, the final protocol action can revert if the destination state changes while funds are in transit. The bridging is complete, but the intended outcome is not. Users end up holding the asset on the destination chain rather than completing the intended action.

This limitation becomes clearer when compared to aggregation in Web2. According to Ben Thompson’s Aggregation Theory, aggregation works because the aggregator controls fulfilment. Marketplaces and platforms not only surface options, but also standardise how transactions are fulfilled. Even when supply is fragmented, execution remains predictable because the aggregator enforces it.

Crypto aggregation operates differently. Aggregators can suggest paths, but execution is carried out by independent protocols deployed across different chains. Increased usage doesn’t give the aggregator more control or improve reliability. It simply increases the number of times users are exposed to inconsistent execution outcomes.

This is why crypto aggregators do not produce strong demand-side lock-in. Users continue to switch based on incentives, price, and liquidity because aggregation alone does not guarantee the outcome users want.

Aggregation without a bias for action works up to a point. Once information is available everywhere, what users need is not more options but an infrastructure that can complete their desired actions.

From Transactions to Intents

Information is no longer the bottleneck. What matters now is how users’ choices become on-chain instructions.

In practice, this means users submit a blockchain transaction.

A transaction specifies exactly which contracts to call, in what order, with what parameters and fee conditions. The network executes it as written or rejects it if any condition fails.

This is important because execution conditions on a blockchain are not static.

Liquidity changes from block to block as trades settle, pools rebalance, and gas prices move continuously based on network demand. Cross-chain transfers introduce additional delays that are beyond the control of any single chain. The transaction ordering can change at any time, depending on how sequencers or validators include transactions in a block.

In a multi-chain system, these factors vary not only over time but also across environments. If conditions change between submission and execution, as they often do, the transaction cannot adapt. It executes under the assumptions encoded at submission time, or it fails. The result is that either users face failed transactions or worse pricing. Neither is tolerable when they happen often.

The constraint is not that commitment exists, but when it happens. In a standard transaction, the user has to commit very early to a specific path. That path is fixed even before the system has access to the most current state.

Avoiding premature commitment requires a different kind of input. Instead of submitting an instruction that describes how execution should happen, the user submits a statement describing the outcome they want to achieve, along with the conditions under which that outcome is acceptable. The system is not told which contracts to call or which route to follow. It is told what must be true when execution completes.

For example, a user can instruct “acquire X amount of an asset within a given price range before a deadline” without specifying where liquidity should come from or which chain to use.

These instructions are called “intents”.

Pre-commitment still exists. A specific path must still be chosen before the transaction settles. The difference is when that choice is made. With intents, the path is determined closer to execution, using live information about liquidity, fees, and network conditions.

This separation between what the user wants and how execution occurs allows specialist actors, called solvers, to take over. When a user states an intent, they enter a shared execution environment where solvers observe it, simulate possible outcomes against live state, and compete to fulfil it under the specified constraints. The solver that delivers the best valid outcome earns a fee for the service.

What changes in an intent-based model is where most of the path-selection risk sits. Instead of users committing to a fixed set of steps upfront, solvers determine whether and how an intent can be completed under current conditions. If liquidity shifts or prices move, the solver must either adapt the path in real time or decline to fill the intent.

This reduces the user’s exposure to stale execution assumptions, but does not eliminate risk. Users still depend on competitive solvers and healthy market conditions. If there is no demand or liquidity for the desired asset, the intent may not be fulfilled.

Solvers compete only in markets where they can make a profit. When a user wants to participate in a market that’s not profitable for solvers, they may still be unable to sell the asset at the desired price.

Intents do not solve the coordination problem on their own. Solvers still need infrastructure that can reach liquidity across fragmented environments and adapt in real time. The open question is who builds the infrastructure that makes fulfilment dependable at scale.

Interoperable Markets

Early DeFi was built around the idea of “money legos”. Protocols were designed to be composable, meaning you could take an asset from one protocol, use it in another, and build increasingly complex financial products by stacking these primitives together. However, this only worked as long as everything lived on the same chain.

When execution spread across chains, composability broke. The building blocks still existed, but no longer snapped together. Every interaction now required assumptions about which chain, bridge, and pool would hold up under stress.

This kind of fragmentation is not unique to crypto. It’s similar to what Plaid solved for applications that wanted to read financial data across multiple banks.

Plaid created a common interface that allowed applications to read financial data regardless of where it resided. Developers could assume that balances and transaction history are accessible and focus on designing experiences rather than rebuilding plumbing.

Before LI.FI, liquidity technically existed across chains, but applications could not assume that access to it would be consistent. Every bridge, DEX, and chain exposes liquidity through its own interface, with its own limits. If an application wanted to move value across chains or compose multiple actions, it had to manage those differences. Building anything cross-chain meant stitching together bridges, swaps, gas handling, and fallback logic, and then hoping execution held up under real-world conditions.

LI.FI applies the same pattern to liquidity. It standardises how liquidity is accessed and executed across them. Applications no longer need to care where liquidity lives, which bridge supports which asset, or how execution should be sequenced. They can treat liquidity as something that can be reached programmatically and reliably, regardless of the underlying path.

The complexity now moves into the infrastructure layer, where it can be managed once rather than being rebuilt repeatedly. Instead of determining where liquidity lives, what matters now is who decides how it gets used.

At that point, the scarce resource is not liquidity itself, but rather order flow. In traditional finance, the two are largely interchangeable, but the subtle differences matter in crypto. Liquidity is passive. It sits in pools, vaults, and contracts. Order flow is active. It represents users and applications trying to swap, move, rebalance, or deploy capital. A pool with $10 million in USDC on a chain where nothing is happening has liquidity but no order flow.

What changes the equation is consistent, programmable access. When an orchestration layer reliably routes activity to the venues where execution is best, two things happen. First, market makers, solvers, and liquidity providers allocate to surfaces where order flow already concentrates and outcomes are reliable. Second, the system that controls the routing gains leverage. It can set expectations for execution quality, enforce latency and reliability requirements, and, over time, shape where liquidity is deployed.

This is the flywheel LI.FI is building. More order flow attracts more competitive participants. More competition improves execution. Better execution attracts more integrations, which generates more order flow.

Interoperability concentrates order flow, and concentrated order flow becomes a marketplace.

When an interoperability layer aggregates meaningful order flow, it starts behaving like a marketplace by default. The job stops being “move tokens between chains” and becomes “take a user’s intent and reliably fulfil it under live conditions”. The user expectation has gone from “move tokens between chains” to “perform this action with these constraints”. In response, the interoperability layer must evolve as well. As a result, interoperability has order flow transitioning towards a marketplace.

Meta aggregation is a marketplace where intents serve as demand and solvers as supply. Multiple executors compete to fulfil each intent. The system picks the best option given constraints such as price, speed, reliability, and compliance.

Solvers that fail, misquote, or deliver poor outcomes lose flow, margin, and eventually relevance.

For solvers and liquidity providers, the value is predictable flow. Predictable flow lets them quote tighter, rebalance more efficiently, and run thinner margins. Unpredictable flow forces them to hold idle inventory, price wider, and avoid long-tail venues.

That supply and demand loop is also why marketplaces are hard to bootstrap. Nobody wants to supply early when demand is uncertain, and demand does not show up until outcomes are reliable. The typical solution is to proxy one side until the other has a reason to participate.

How LI.FI Became a Marketplace

LI.FI took a path that evolved into a marketplace. It started as distribution and execution aggregation.

The first move was embedding into production applications. Over 800 businesses now route through LI.FI. The demand side of the marketplace was seeded before the supply side existed.

The second move was going beyond comparing prices to handling the actual execution. LI.FI deployed smart contracts across chains that could combine multiple steps into a single transaction for the user. Say you want to swap a token on Arbitrum, bridge it to Base, and swap it again into a different token there. Without LI.FI, that is three separate transactions you sign and monitor.

With LI.FI’s contracts in the middle, the user signs once. This entire process is non-custodial. LIFI doesn’t get to hold or direct users’ funds at any point. Funds move directly between smart contracts while remaining under user control at all times. The contract handles the swap, sends it through the bridge, picks it up on the other side, completes the final swap, and delivers the result.

The system that makes this possible uses a modular architecture based on the Diamond Proxy pattern, also known as ERC-2535. Applications interact with a single entry point to the contract, which delegates actions to specialised modules based on what needs to happen. Each module handles one capability, such as a specific bridge, a DEX, an approval, or a protocol deposit. These capabilities can be added, replaced, or updated without redeploying the rest of the system.

As the cross-chain landscape is constantly evolving, this makes sense. New bridges launch, existing ones deprecate routes, DEX aggregators update interfaces, and protocols change their contract logic. A monolithic contract architecture accumulates technical debt with every change. Each new bridge integration or DEX update touches shared code paths, which means every modification carries the risk of breaking something that was already working.

Over time, engineering teams spend more effort maintaining existing integrations than building new ones. The Diamond Proxy pattern avoids this by keeping each integration isolated. When a bridge updates its interface, LI.FI updates one module. When a new chain launches, LI.FI deploys a new module to the entry point. The rest of the system is untouched.

For simpler integrations, LI.FI forwards generic contract calls with encoded data. For most major protocols, however, it maintains dedicated execution modules that interact directly with protocol interfaces. As of early 2026, LI.FI supports more than 80 such modules covering bridges, DEXs, and DeFi protocols. Each module is a well-defined execution primitive that integrators can combine in different ways.

This architecture also enabled LI.FI to offer redundancy at the infrastructure level. Some integrations intentionally overlap to ensure that if one path fails, another is available. Enterprise clients like Robinhood and Binance integrate LI.FI because they need consistent execution across every route, on every chain, with fallback mechanisms built in. The Diamond Proxy is what makes that possible without the contract system becoming unmanageable.

The practical result was that LI.FI could observe the full lifecycle of every transaction passing through its contracts. Where did failures happen? Which bridges were slow on which routes? Which DEX aggregators won on which chains? This data telemetry became a compounding advantage, one that competitors could only replicate by processing comparable volume through comparable infrastructure.

The third move was building a consumer surface. Jumper is LI.FI’s retail interface, where UX primitives ship fast and inform the B2B product. The clearest example is destination gas. Jumper’s Refuel feature lets users acquire native gas on the destination chain during the journey, eliminating the “funds arrived but I cannot do anything” failure mode. Once you stop failing at infrastructure, the question becomes whether the user achieved their intended outcome. Did the swap execute at a fair price? That is the problem worth optimising for.

The fourth move was making journeys programmable. Pure aggregation hits a scaling wall once users start expressing multi-step outcomes instead of simple “bridge and swap” requests. With a few chains, bespoke integrations are manageable. However, every new chain introduces new edge cases, every protocol has different smart contract interfaces, and composing actions across them manually does not scale. LI.FI spent two years building Composer to solve this.

Composer is a transaction orchestration primitive. It lets developers define multi-step, multi-chain DeFi workflows and execute them as a single transaction. The user signs once on their source chain. Everything else happens behind the scenes.

Under the hood, Composer compiles user journeys into executable blockchain instructions using a domain-specific language in TypeScript. A developer can write logic such as check the wallet balance, if it exceeds $1,000, swap into ETH, deposit the ETH as collateral into Aave, take out a 75% loan, and swap the proceeds into a stablecoin: multi-chain, multi-protocol, one transaction. The system uses dynamic call data injection to pass outputs from each step into the next, so there is no manual parameter handling between legs of the journey.

Critically, Composer includes its own transaction simulator. Before anything touches the blockchain, the full execution path is simulated to verify it will work. This is more than a nice-to-have feature. Failed transactions on complex multi-step journeys cost users gas and erode trust. Pre-execution simulation turns “hope it works” into “verified it will work”.

The practical result is visible in products like Jumper Earn, where users make one-click cross-chain vault deposits. Behind that single click, Composer is orchestrating a swap, a bridge, and a protocol deposit across different chains. It is also what LI.FI is rolling out with enterprise partners like Alipay, where the end user cares only about the outcome (yield) and not the six on-chain steps required to get there.

Composer matters because it expands what applications can express and execute with a single integration.

Before Composer, an intent was limited to “swap this” or “bridge that”. After Composer, an intent can describe an entire strategy. When you can express an outcome programmatically and simulate it cheaply, you can auction the right to fulfil it.

The fifth move was opening execution into an auction. LI.FI’s intent system is a modular auction layer where solvers compete to fulfil user intents. As mentioned earlier, the system is non-custodial. User funds remain in their wallets and move only via transactions they explicitly sign.

Every intent system faces a bootstrap problem: solvers won’t appear without order flow, and order flow needs solvers. In LI.FI’s case, years of aggregation had already built the demand side. Billions in monthly volume flowing through MetaMask, Phantom, Robinhood, and hundreds of partners meant solvers had an economic reason to participate from day one. The cold start problem was already solved before the product launched by proxying demand.

That is why acquiring Catalyst made strategic sense. The logic was to buy the best execution technology and pair it with the order flow moat that no intent-native competitor could replicate. Catalyst brought a flexible architecture supporting multiple transaction flows, settlement layers, and oracle options. LI.FI brought the distribution that makes solver economics work.

The intent system also changes what developers can build. A lending protocol cares about collateral thresholds and liquidation timing. A wallet cares about speed and simplicity. A payments app cares about predictable fees and settlement finality. Each has a different definition of “good execution”, and a generic intent format would be a terrible design choice to serve all of them.

LI.FI’s architecture lets each integrator define application-specific intents with their own constraints, routing preferences, and execution criteria, but all of this is resolved through the same solver marketplace. Developers get the expressiveness to build complex cross-chain flows through a single integration, without having to reason about the underlying execution paths themselves.

The result is an intent system that did not have to compromise on either side. LI.FI still aggregates other intent systems and does not lock applications into a single execution path. However, it now also operates its own solver marketplace, where the same order flow concentration that made aggregation defensible makes intent execution viable.

Google’s ad exchange followed a similar journey. Before programmatic advertising, advertisers negotiated with individual publishers. Google built an orchestration layer that aggregated both sides and introduced real-time bidding. The exchange became the marketplace for price discovery, its leverage stemming from concentrated order flow across fragmented supply.

The economics were similar, too. Advertisers did not want to integrate with thousands of publishers individually, nor did publishers want to integrate with advertisers. Both sides were okay with paying the exchange fees because the alternative was worse. The exchange captured value by reducing friction and allocating attention efficiently.

LI.FI sits in the same position. Applications do not want to integrate with dozens of bridges and DEXs individually. Execution providers do not want to build relationships with thousands of applications. Both sides accept routing through LI.FI because the alternative is worse.

Why apps keep outsourcing to LI.FI

A lending market, perp DEX, or wallet competes on liquidity, spreads, risk, and distribution. Interop reliability sits in the critical path, but it is still a cost centre. Owning the entire execution surface across chains, bridges, and protocols means running an integration business.

Many teams integrate with LI.FI because it lets them buy a single execution surface that already sits on top of many venues, learns from aggregate telemetry, and is moving towards outcome-based fulfilment via solver competition. As more apps route deposits and complex actions through the same layer, order flow concentrates further. That concentration is the moat. It improves solver economics, enhances routing quality, and deepens the marketplace.

The game-theoretical outcome is that one player concentrates order flow, and solvers have to compete for access to it. If they try to extract too much margin, someone else undercuts them. If they fail to execute, they lose allocation. The dominant order flow aggregator becomes the default price-setting venue.

Why Now?

When everything is growing, even mediocre infrastructure looks like it’s working. That’s what happened in the early days of crypto interop. But it’s changing now. The bar has moved from “does it connect chains” to “does the transaction actually go through every time”. Teams that aren’t specialised enough to win on performance or integrated enough to win on distribution are getting squeezed out. This consolidation is happening alongside three shifts that make orchestration infrastructure more necessary than ever.

1. Fragmentation is compounding.

The early fragmentation problem was about chains. That problem still exists, but it is now joined by asset-level and rail-level fragmentation. Stablecoins illustrate this clearly because they are the single most used asset class in crypto, and the fragmentation runs through every layer. Stablecoins are fragmented at three levels – issuer, chain, and token standard.

Starting with the number of instruments, the stablecoin market now exceeds $300 billion in total supply, up from roughly $205 billion at the start of 2025. Although Tether and Circle control ~85% of the market share, there are more than 50 issuers with over $100 million in supply. These stablecoins are scattered throughout applications on multiple chains. The chart below shows the distribution of stablecoins across applications.

USDC is natively issued on 30 networks. USDT is supported on roughly 20 chains, and with USDT0 (its omnichain variant on LayerZero’s OFT standard), it reaches 18 more. DefiLlama tracks stablecoin activity across more than 100 chains.

Finally, there is the question of how the same stablecoin is represented. USDT exists as an ERC-20 token on Ethereum, a TRC-20 token on Tron, a BEP-20 token on BNB Chain, an SPL token on Solana, and so on. These are not interchangeable. You cannot send TRC-20 USDT to an ERC-20 wallet. Beyond chain-native standards, there is now a competing layer of cross-chain token standards designed to unify exactly this fragmentation.

LayerZero’s OFT (Omnichain Fungible Token) standard powers USDT0, USDe, PYUSD, and others. Wormhole offers NTT (Native Token Transfer). Chainlink has CCT (Cross-Chain Token). Axelar has ITS (Inter-chain Token Service). A newer standard, ERC-7802, allows tokens to expose structured metadata about their origin chain and verification endpoints. Each standard has its own set of stablecoin integrations, security assumptions, and developer ecosystem. The infrastructure designed to solve fragmentation is itself fragmenting.

The result is that a single dollar-pegged asset can exist as native USDT on Ethereum, native USDT on Tron, USDT0 via LayerZero’s OFT, wrapped USDT through various bridges, and bridged representations through A single dollar-pegged asset can simultaneously exist as native USDT on Ethereum, USDT0 via LayerZero, wrapped USDT through various bridges, and bridged representations through Wormhole or Chainlink. The same underlying dollar had different contracts, pools, and trust models. The coordination problem is now stablecoins × chains × token standards, and every combination creates a distinct execution surface.

As more real-world assets get tokenised, the same pattern compounds. A US Treasury bond might exist as an OFT on one chain, a native mint on another, and a CCT on a third. These differences determine how an asset is traded, bridged, and composed.

The execution surface is also splitting. Compliant rails with KYC’d liquidity pools are emerging for institutions. Privacy rails are emerging for those who need to keep the data confidential. Every new rail type is an additional execution surface that needs to be aggregated.

Companies are building their own chains to retain control over compliance, fee structures, and value accrual. Robinhood has its own chain. Stripe acquired Bridge and is building Tempo. Circle launched Arc. These are distribution-motivated chains built by companies with hundreds of millions of existing users, and every one of them creates new liquidity venues, new token representations, and new routing complexity.

If tokenisation follows the trajectory its proponents expect, every value chain that moves on-chain will introduce its own standards, its own chains, and its own fragmentation. The coordination problem will compound before it converges.

2. Regulation is accelerating asset issuance.

The GENIUS Act, now law, established the first comprehensive federal stablecoin framework in the United States. Although the regulatory clarity is good for adoption, the second-order effect is that it legitimised stablecoins as a product class and opened the door for a wave of new entrants.

This is not just a US phenomenon. MiCA is in force in Europe. Hong Kong passed its Stablecoin Ordinance in 2025. Each jurisdiction has its own compliance requirements, approved issuers, and preferred rails.

The GENIUS Act also prohibits stablecoin issuers from paying interest to holders. It does two things.

1. It drives the growth of crypto native stablecoins like USDe and USDH that are willing to pass on the yield to users.

2. It pushes yield-seeking innovation into adjacent instruments: tokenised treasuries and structured products that sit one layer above the stablecoin itself.

Each of these instruments needs its own token representation, its own chain deployments, and its own liquidity. Regulatory legitimacy accelerates issuance. Issuance accelerates fragmentation, and fragmentation accelerates the need for a coordinated infrastructure that can map, route, and execute across it all.

3. More and more executions are becoming autonomous.

As more AI-driven and agentic systems come online, execution increasingly occurs without a human in the loop. Agents rebalance portfolios, deploy capital, and move funds in accordance with predefined rules. There is no user selecting chains, bridges, or routes. There is only an intent and a requirement that it executes correctly under changing conditions.

According to Flashbots, autonomous execution is increasingly a first-order driver of how blockspace is consumed and how order flow is routed. Spam bots across rollups consume more than 50% of the blockspace while paying less than 10% fees. Meanwhile, private order flow accounted for over half of Ethereum gas usage. Bots on Ethereum accounted for more than 73% of DEX activity in August 2025, reinforcing the idea that execution quality and inclusion guarantees are increasingly determined upstream of the chain itself.

Agents do not encode chain-specific logic, bridge choices, or sequencing decisions. They express outcomes. They need something downstream to resolve their intent into an executable path, using live information about liquidity, fees, and network conditions. As the agentic economy grows, systems that can reliably execute without human intervention will matter more than interfaces.

The coordination problem is permanent. There will be no single chain; there will be hundreds, each with economic and technical reasons to exist. The infrastructure that maps standards, routes value, and coordinates execution across all of them becomes increasingly essential with every chain that launches.

What Could Undo This?

LI.FI’s thesis depends on fragmentation being a lasting feature of crypto, not a temporary phase.

The first is chain convergence. If activity consolidates onto one or two chains to the point where cross-chain movement becomes irrelevant, there is not much left to orchestrate. If everything that matters happens on Hyperliquid, you do not need a meta-aggregator. The counterargument is that if this were to happen, it would have by now. There have been stretches when one ecosystem dominated activity, but none in which the others disappeared. Cardano still exists. Every cycle produces new chains with real usage, real liquidity, and real reasons to exist.

The second scenario is standard convergence. Currently, the same asset can be represented across multiple cross-chain token standards, such as OFT, CCTP, NTT, and others. If one of these standards wins so completely that everything runs on the same rails, the mapping and routing complexity that LI.FI navigates gets much simpler. Today, at least three standards are prominent and growing independently. If a single standard captures enough adoption, the value of sitting between them shrinks.

Neither outcome looks likely in the near term, but they represent the structural conditions under which the LI.FI thesis would need to be revisited.

Where Interoperability Goes From Here

If the Internet pattern holds, raw transport is on the path to becoming a utility. Pure bridges are already headed in that direction. For a $1,000 cross-chain swap, most bridges and aggregators will land you at a similar execution price. The meaningful difference is whether the solution can navigate the maze of DEXs and bridges when some stop working, route around failures without the user noticing, and do so consistently across dozens of chains. Applications like wallets integrating cross-chain functionality natively are not necessarily optimising for the cheapest option, but for the one that does not break.

As this happens, bridges or DEXs will lose their identity. They fade into the background, as physical network infrastructure did once the Internet standardised its transport layer. Users and applications will stop thinking about which bridge moves their assets. They will care about whether assets arrive, and at what cost. In this world, the value migrates up the stack, away from the pipes and towards whoever coordinates the journey between what a user wants and where liquidity actually sits.

The solver networks that attract the most order flow will attract the best solvers. The best solvers will deliver better execution. Better execution will attract more order flow. The flywheel is already observable. Only consistent demand makes this durable, though. Orchestration solutions are in a unique position as they sit close to where the demand originates.

Bridges and liquidity venues hold supply; orchestration layers control demand routing. As order flow concentrates through these layers, suppliers compete for access to the demand surface rather than the other way around. Greater competition means their margins compress. Their negotiating leverage diminishes. The value accrues to whoever sits between user intent and liquidity deployment.

Then the cycle restarts at a higher layer of abstraction.

When execution infrastructure is reliable enough that applications stop worrying about whether cross-chain transactions land, they start worrying about how those transactions land.

A payments company needs predictable fees and settlement times. A regulated exchange needs KYB-verified execution partners. An AI agent rebalancing a portfolio needs reliability guarantees that differ from those required by a human clicking “swap.” Alipay does not care about the composable transaction steps underneath. They care about one-click yield. A company tokenising US Treasuries needs solvers with permissioned access to mint and burn those specific tokens, not a generic solver pool.

This is the fragmentation that comes after connectivity is solved. First, bridges enabled the movement of assets across chains. Then, token standards from the likes of LayerZero, Chainlink, and Circle made interoperability more structured. At the application layer, users’ needs diverge faster than any generalised system can accommodate. Infrastructure that forces uniformity loses to infrastructure that enables specificity. The teams that win will be the ones letting applications define their own constraints, solver selection, oracle preferences, compliance filters, and execution timing, without having to rebuild the stack from scratch.

There is no single end state here. We will be tokenising different assets using different standards across different chains for the foreseeable future. The coordination problem will compound before it converges.

LI.FI’s bet is that the on-chain world is headed towards a landscape of millions of tokens, hundreds of standards, hundreds of chains that matter, and application-specific intents that a one-size-fits-all execution cannot serve. In that world, the durable advantage belongs to whoever has the deepest integration footprint into the surfaces where intent originates, the solver depth to execute across the widest range of conditions, and the flexibility to let each application define what “good execution” means for its users.

Strowger did not build a better telephone operator. He made the operator unnecessary by letting the caller’s intent route itself. The bridges, DEXs, and liquidity venues are the operators. LI.FI is building the switch.

Streaming the T20 World Cup,
Saurabh Deshpande

Disclosure: This piece was part of a sponsored collaboration. Authors or DCo may own assets mentioned in the article.

In the previous episode, we explored how LI.FI is tackling fragmentation in a multichain world. This time, we zoom out and ask what happens when crypto’s favourite derivative meets the $100 trillion traditional futures market?

Spotify

iTunes

Robert Shiller published a paper proposing perpetual futures as a way to create derivative markets for illiquid assets like housing back in 1993. The idea sat in academic journals for nearly two decades. Traditional exchanges never picked it up. About two decades later, BitMEX and a group of crypto exchanges turned the perpetuals into an actual product that people could trade.

It caught on because perpetuals solved real problems. People did not have to bother with monthly rollovers or the logistical overhead of physical deliveries. Funding rates kept prices anchored to spot. Within a few years, perpetuals accounted for over 90% of all crypto derivatives volume. The concept that traditional finance invented became crypto’s signature instrument.

Brett Harrison, founder of Architect and its exchange AX, is running the playbook in reverse. AX is a regulated centralised exchange offering perpetual futures on traditional asset classes: gold, silver, currencies like the Indian rupee and Mexican peso, single stocks like Apple and Tesla, and GPU compute power. The settlement rails are crypto native. Customers can post USDC as collateral and move money in seconds rather than waiting days for a wire to clear.

You might wonder why incumbents like CME never did this themselves. The answer lies in regulation, infrastructure and incentives.

US regulation does not clearly classify perpetuals as futures. They might be swaps, which would restrict retail access entirely. CME’s clearing architecture runs on fixed daily settlement cycles incompatible with continuous funding rate mechanisms. And CME earns substantial revenue from customers rolling expiring futures every month and every quarter. Each roll creates two transactions, and both are charged. Launching a superior product that cannibalises that revenue is a textbook innovator’s dilemma.

Traditional futures still win in specific contexts: regulatory clarity under CFTC oversight, central clearing that eliminates counterparty risk, and predictable costs without ongoing funding payments. $19 billion was lost collectively during the 10/10 liquidation event across crypto markets. A lot of it could have been avoided if the exchanges had safeguards like price bands, circuit breakers, and margin calls rather than automatic liquidations. Traditional finance built these failsafes over decades, and AX incorporates them all.

The bet Harrison is making is that for a very large class of participants, from CFD brokers hedging exposure to crypto native firms branching into equities, perpetuals on crypto rails are simply better. CME alone supports more than $8T in average notional daily volume compared to all of crypto at $250B. The overall futures market is estimated to turn over at least $15T per day, and AX wants a piece of that.

It is clear to me that the whole of finance will meaningfully change in a few years, and crypto rails will play a major role. Traditional assets coming on-chain and traditional finance using crypto rails to cut costs and time will likely be a secular trend going forward.

Thinking about new markets,
Saurabh Deshpande

Spotify

iTunes

Imagine a future where JP Morgan settles trillions of dollars on-chain and none of the DeFi behemoths we know today benefit from it. Adoption of the industry does not directly translate to existing ventures benefiting from it. This is why LI.FI is betting on markets, rather than chains. With billions in volume already facilitated, they know a thing or two about how space evolves.

Founder and CEO at LI.FI, Philipp Zentner, joined me for a conversation to unpack where crypto, tokenisation, interoperability, and distribution are headed.

Two major trends are underway in crypto right now.

1. On the one hand, crypto-native players like Trade.xyz, Markets.xyz, and many others are bringing traditional assets to crypto.

2. On the other hand, institutions are using private (customisable) blockchain rails like Canton to tokenise traditional assets.

Amidst these, what is clear is that assets of interest on-chain will increase. Obviously, these will be across different chains and standards, making sure that the fragmentation of assets only increases for the foreseeable future. Meanwhile, it is also true that users don’t want to bother with these standards. They want to go from asset A to asset B in a click, regardless of which chain either lives on, which token standard wraps them, or which compliance framework governs them. The gap between how assets are issued and how users expect to interact with them is growing. Whoever bridges this gap will likely capture the flow of assets.

This is where distribution becomes crucial. Tokenising an asset is only the beginning of the journey. Finding demand for it, getting it into wallets, making it tradeable with tight spreads and reliable execution across venues, that is the hard part. And it is the part most institutions have not yet figured out. LI.FI, with integrations across 800+ applications, from MetaMask to Phantom to Robinhood, LI.FI is positioning itself as the solution to a question institutions are only beginning to ask.

Among so many interop solutions, how is LI.FI building its moat? The answer lies in data telemetry and orderflow. Given LI.FI’s distribution, more and more apps are integrating it. This leads to more orderflow going through LI.FI’s system. All of this data going through LI.FI allows it to understand where the pain points are and make usre that it is the first to address those. This leads to more orderflow, and so on.

What’s interesting to me is how LI.FI, as an infrastructure layer, may have a better understanding of capital flow than the individual apps through which this capital passes. Philipp and his team think of these transactions as a database engineer thinks of queries. Every failed transaction is a data point that makes the next transaction smarter. The conversation delves into why the on-chain world will remain fragmented, and how LI.FI is building to thrive in that fragmented future.

See you in the next one!
Saurabh Deshpande

We closed 2025 with Mike from Talus Network on why blockchains need brains and hands to participate in an agentic economy. We kick off this year asking how a joke has survived twelve years in an industry that forgets everything in twelve days.

Spotify

iTunes

Belief is the true asset. What people believe in, they hold - and it tends to rally in price due to limited supply. Gold is what people are betting on these days. It is up 68% against the dollar in the past year, so anyone holding it is having a good time. Meme assets represent the other end of the belief-monetary value spectrum. They have nothing backing them but a high degree of holders believing in them. In turn, over time, they become money.

Don’t worry - we are not about to shill a cult to you.

We are, however, about to talk about Dogecoin. Crypto’s original meme asset. It was launched as a joke in late 2013 by Billy Markus and Jackson Palmer.

Timothy Stebbing, CTO of House of Doge and Director at the Dogecoin Foundation, joined me for the first episode of the year. In his spare time, he tells people to stay away from crypto as an investment. Instead of investing in Dogecoin, he wants you to spend it. Five billion new coins get minted every year. This mechanism was created as a joke to mock bitcoin maximalists. Turns out roughly 2% inflation is actually pretty good for something you want people to use as currency rather than hoard as digital gold.

But is that what holders are doing? It depends on who you ask. Dogecoin has historically been a retail participant’s preferred “entry-drug” to crypto. It used to be much cheaper. So, tossing less than $1 to get a large number of tokens would appeal to people. It has been exceedingly volatile when compared to its larger cousins - Bitcoin, Ethereum and Solana. Most importantly, it is the only meme asset that has lived through multiple cycles and set a new market-cap high in each cycle. Bitcoin’s distant, amusing cousin somehow teaches retail about crypto better than most VC-backed tokens.

But Dogecoin is a meme that is building something productive. Companies with treasury strategies are buying in. The community now spans memecoin traders, open source developers, and institutional investors. When Metallica performs with a symphony orchestra, two kinds of crowds are in the audience. One full of fans in ripped jeans and Ride the Lightning shirts. The other in suits, the usual symphony crowd. But they were all going to see the same show. That is the Dogecoin community now. People who look different, speak different languages, come from different worlds. But something made them all show up.

Doge worked and perhaps outperformed many of its peers, probably because it set out without intending to. It is an idea that took its time and played out over a decade. Most crypto-native assets cannot make that claim. There is no legacy of sponsoring NASCAR drivers in the trenches of PumpFun today. Perhaps that is what will be needed for those assets to catch up with Doge. We don’t quite know. But here’s what’s evident: Doge has survived a decade, and it may survive another one. Tune in to our conversation with Timothy to understand the reasons why.

Trying to do only good every day,
Saurabh Deshpande

RFQ, aka Request for Quote, offers a simple premise. What if you could simply ask for a price before you bought? What if that price was guaranteed, signed, and settled without anyone else getting a peek? Bebop has been dominating the category for the past two years. These primitives are key to scaling new markets and have been used in TradFi for decades. Bebop has been recreating it on-chain to aid developers looking to empower users to access liquidity across asset types. They do $500 million a week in volume, according to DefiLama.

Today’s article, written in collaboration with Katia Banina, looks at how an ancient practice is being rebuilt on-chain, and why Bebop thinks it has found a way to filter out the lemons.

If you are exploring new markets, providing liquidity, or are a market-maker firm, consider reaching out to us at venture@decentralised.co. We are scouting for partners to provide liquidity to several of our portfolio companies.

Information about the product you purchase is critical to establish its price. The more you know before making a purchase, the easier the price discovery process is. Often, you verify the quality of a product only after you have purchased it. If you can verify it cheaply and have some recourse in the form of return or refund, it is all good. During the summer, I purchase mangoes directly from the farm and pay a premium over the market price. The premium is worth it due to the superior quality of these mangoes.

The farmer knows they are high-quality mangoes; I can verify it against regular mangoes. And agree to pay the premium. The information asymmetry between the seller and buyer disappears when verification is inexpensive.

What happens when the ability to verify is no longer available?

In 1970, George Akerlof wrote a paper called The Market for ‘Lemons’, for which he would win the Nobel prize in economics in 2001. The paper refers to the idea that information asymmetry destroys markets. He uses an example of used cars. Only the sellers know whether their cars run smoothly. The buyers do not, until they’ve bought and used these cars. Buyers anticipate this information asymmetry, and they tend to pay average prices. Owners of good cars (peaches) cannot afford to sell at average prices, and they eventually exit the market. So, all that is left are the low-quality cars and the lemons.

When three market forces align i) the ability to profit from an event, ii) access to technology to exploit it, and iii) access to material information, sophisticated market participants will move to maximise their profits at the expense of average participants. Call it the “three-body problem” of markets: when these three forces interact, they create conditions where extraction becomes inevitable.

You do not want that. Regulators and watchdogs are supposed to monitor markets to prevent these three forces from aligning in ways that harm regular participants.

We aimed to address these issues arising from information asymmetry by utilising the Web3 infrastructure. Blockchains are public. Smart contracts are auditable. Every transaction is visible. The code is law, and the law is transparent. But it introduced different issues.

Traditional finance exchanges are based on order books. They are populated with buyers’ bids and sellers’ asks. When you market buy something, you fill the lowest ask first and then the second lowest, and so on. And the other way around when you market sell. However, given Ethereum’s slow blocks and gas limit constraints, placing and changing limit orders was difficult, making on-chain order book-based designs infeasible. AMMs came to the rescue. It also meant that nobody had to keep placing limit orders. You had a passive pool of capital that people could trade against.

AMMs allowed market participants to buy and sell assets based on a formula that prevented prices from fluctuating excessively. Importantly, that could operate independently from any single party that is opinionated on the fair asset value, i.e., a market maker. This is what you want from a market. Buyers and sellers should be able to trade in reasonably high quantities without significantly affecting the price.

However, the issue arises because of the way transactions are ordered from the mempool. When you want to buy or sell an asset and broadcast the order, everyone can see it. If there’s a possibility of ordering transactions in a certain way to profit, someone will do it. And it is a zero-sum game, when someone profits, someone is losing, which is most likely a retail participant. There’s the issue of how sophisticated players can exploit participants who lack information or access to technology. Sumanth has written extensively about it in his discussion of the inevitability of MEV.

When no professional market makers would touch crypto, automated pools allowed anyone to provide liquidity and earn fees. They made DeFi possible. But as markets mature, different tools emerge for different needs.

RFQ, also known as a Request for Quote, is an ancient concept. Humans have done it since the earliest days of commerce. It is about asking around for the best price before making a large purchase commitment. It then became a practice in industrial procurement. When governments award construction contracts to companies or an automaker like Tesla or General Motors wants to procure 5,000 tonnes of steel, they send specific requests to the relevant parties to obtain the best offer.

RFQ found its way to finance in foreign exchange, fixed income, and wider financial markets that are traded over the counter (OTC). Instead of posting orders to the public orderbook, you ask for quotes from market makers, who provide guaranteed prices for short periods of time.

The crux here is that there is competition to fill an order, allowing the buyer to get a better price. Consistent better pricing gives buyers the confidence, and thus, demand remains robust. Competition leads to better price discovery.

Financial markets adopted RFQ for the same reasons. It offers better UX. For example, Robinhood uses quotes from market makers for asset prices instead of routing to exchanges. In cases when orders are large, customised, or in markets with thin public order books or markets like foreign exchange with no order books, you can get a significantly worse price that eats into your margins.

A company that needs to exchange $100 million would not want to move the market by trading it in public markets. They would request quotes from several banks. Banks would have their internal risk assessment frameworks to provide quotes that stand for a stipulated time.

RFQ brings several inherent advantages that matter increasingly as trading becomes more sophisticated:

Superior price discovery. Market makers have a holistic view of the entire market. They see prices across centralised exchanges, multiple DEXs, derivatives markets, and order flow patterns. A Uniswap pool knows only what’s happened within that specific pool. When you request a quote, the market maker synthesises information from every venue they monitor, giving you pricing that reflects the genuine market rather than the isolated state of one pool.

CEX-quality execution on-chain. The best market makers in crypto are the same firms providing liquidity on Binance, Coinbase, and OKX. RFQ allows them to bring that same quality of pricing and execution to on-chain trading. You get centralised exchange depth (or better, because capital can sit on the same chain and be utilised across venues) without the custody risk.

Dramatic capital efficiency. In an AMM, liquidity must sit idle in pools, waiting for trades. RFQ liquidity lives in the market maker’s inventory system, where it can be actively managed, hedged across venues, and deployed only when needed. This means market makers can offer much larger fills at competitive prices. Where a pool might require $10 million of TVL to fill a $100,000 trade without excessive slippage, an RFQ market maker can fill that same trade using exactly $100,000 because they hedge dynamically across multiple venues.

Guaranteed execution and fixed prices. When a market maker signs a quote, they commit cryptographically. You know exactly what price you will receive. There is no slippage between quote and execution, no price that moves against you while your transaction waits in the mempool. The certainty matters enormously for larger trades or during volatile periods.

Modern intent protocols and sophisticated solvers are also addressing many of the MEV challenges that plagued early AMMs. The infrastructure is getting better across the board. But RFQ offers something fundamentally different: professional market-making rather than formula-based pricing. It’s not that AMMs are broken, it’s that RFQ enables use cases that passive pools cannot serve effectively.

Where AMMs excel at providing baseline liquidity for routine swaps, RFQ shines for larger sizes, exotic pairs, volatile conditions, and any situation where execution certainty matters more than permissionless simplicity. The future likely involves both, with aggregation layers intelligently routing each trade to wherever it can be executed most efficiently.

Both AMMs and RFQs have different privacy characteristics, each with its own tradeoffs.

When you request a quote through RFQ, you reveal your trading intent to the market makers before committing to anything. They see that you want to sell 100 ETH. They know current prices across every venue they monitor, and they understand which direction the market is moving. You have less complete information. This creates an asymmetry.

However, this asymmetry can be managed through good system design. Protocols implement various mechanisms to limit what information flows to makers, when they see it, and how they can respond. The challenge is not inherent to RFQ as a concept—it’s a design problem with known solutions.

Interestingly, AMMs reveal nothing about your intent during the quoting phase because there is no quoting phase. You see the current pool state and decide whether to trade. Your intent only becomes visible when you submit your transaction. At that moment, though, it becomes completely public. Your pending transaction sits in the mempool, where anyone can see it, analyse it, and act on it before you execute.

This is where RFQ provides a crucial advantage. Once you accept a quote, the market maker has cryptographically committed to that price. Your transaction settles atomically with no opportunity for someone else to insert themselves in between. No frontrunning, no sandwiching, no MEV extraction from reordering. The maker sees your intent earlier in the process, but no one can act on it adversarially.

The question becomes: what happens between the moment you request a trade and the moment it executes? In AMM pools broadcasting to the public mempool, sophisticated actors can sandwich your transaction or frontrun it, extracting value from the gap between your expected and actual execution price.

In RFQ systems, the signed quote eliminates this attack surface. The price is fixed at the moment of signature. So, there’s no slippage to exploit or opportunity to insert transactions before or after yours. For small routine swaps, the difference may be negligible. For anything substantial, guaranteed execution at a committed price offers meaningful protection against the extraction strategies that are prevalent in public transaction pools.

We should also note that both models continue to evolve. Intent protocols are adding privacy-preserving mechanisms to AMM execution. RFQ platforms are implementing techniques to minimise what information flows to makers and when. The privacy landscape is not static, and good design can address vulnerabilities in both architectures.

History of on-chain RFQ

Airswap kept RFQ peer-to-peer early in DeFi. Traders found each other, agreed on prices off-chain, and settled on-chain through the Airswap contract. This was suitable for sporadic trades but not for high-frequency trading.

In a traditional RFQ, once a maker quotes you, they are stuck with that price. Airswap implemented LastLook, which flips it. You sign the order, then the maker gets “last look” before executing. This makes trades gasless for users. The maker pays gas and takes execution risk.

0x approached the market from the other side and created an open protocol that anyone could use. Their RFQ-T (RFQ for Traders) system, launched around 2019, allowed professional market makers to provide quotes through a standardised API.

The separation of the infrastructure layer from the application layer allowed market makers to plug into 0x’s system, and any front-end could request quotes from these makers. This worked well for professional trading firms and aggregators, such as Matcha (0x’s own interface), but it remained primarily an infrastructure - most retail users never directly interacted with 0x’s RFQ system.

Every trading venue has a chicken-and-egg problem of liquidity and users. Without liquidity, there will not be users, and without users, there is no liquidity. Hashflow, launched in 2021, addressed the liquidity problem by throwing money at it. But capital alone will not solve the problem if using the product comes with technical overheads. Hashflow developed a pool-based architecture to address the issue. Traditional market makers were figuring out blockchains; you could not just throw smart contracts at them and expect them to solve technical complexities in-house. They could park capital in pools, and Hashflow used it for RFQ.

0x demonstrated that RFQ could work on-chain, but it did not solve the go-to-market challenge. How do you make RFQ accessible and understandable to regular DeFi users who were accustomed to simply swapping on Uniswap?

Hashflow, launched in 2021, took an aggressive growth approach. They raised significant capital and used token incentives to onboard market makers. For a nascent market where order flow was uncertain, these incentives proved effective in bootstrapping liquidity. Hashflow positioned itself as a full-stack RFQ platform with its own interface, market maker network, and user base.

When the incentives were on, Hashflow volumes declined. In 2023, the overall 1inch volume increased around the USDC depeg.

One can think of RFQ as a quote produced from an order book that’s frozen in time. Market maker is essentially pausing time and finding the best quote from across venues. It usually yields better results than trading on a single venue. This can be verified with low slippage and high depth offered by Bebop on many pairs. But RFQ does have its issues with informed toxic flow.

Imagine you are a liquidity provider in a DeFi RFQ system, offering guaranteed price quotes that live for 30 to 60 seconds. Every trade request that comes through could fall anywhere on a spectrum, ranging from toxic to soft on one axis and from uninformed to informed on the other.

In the bottom-right quadrant sits spam. These are high-frequency, uninformed trades that arrive without any real market knowledge. They are soft because they follow predictable patterns and do not adversely select against you.

Move up to the top-right quadrant and you find good traders. These are informed participants who understand markets deeply but are making genuine directional trades. A fund manager rebalancing positions, a treasury executing a planned allocation. Still profitable to serve. But now the trouble begins.

The extractors and MEV searchers lurk in the top-left quadrant. These are highly informed toxic traders who understand precisely what they are doing. They might request two quotes from you 15 seconds apart, then execute both in the same atomic transaction, profiting from any price movement while you absorb the loss. Or they will sit on your quote, simultaneously watching prices across other venues, only executing when they can arbitrage against you.

They might even slice a large order into ten smaller pieces to exploit your tighter spreads on small sizes, then bundle everything into one transaction. They get the price of ten small trades with the economics of one large trade.

You cannot tell these quadrants apart. Every request looks identical on-chain. Just as used car buyers could not distinguish lemons from peaches, you cannot distinguish MEV searchers from good traders or spam until you have been arbitraged. So you widen your spreads to protect yourself. You shorten quote validity windows. You become more cautious. But this defensive posture punishes everyone, including the good traders and even the spam flow that would have been profitable to serve. Other liquidity providers observe the same pattern and follow suit.

Eventually, the best makers exit entirely because they cannot sustainably serve a market where they cannot accurately price risk. The toxic flow has degraded the entire market structure, driving out the quality liquidity providers and leaving behind only those willing to extract value or charge prohibitive spreads. Remember how bad sellers drove buyers out of the market, as seen in the case of used cars? In this case, instead of bad sellers driving out good sellers, bad buyers drive out good sellers.

RFQs protect trades from leakage. Liquidity is created on demand. The taker asks privately, the maker quotes privately, both sides decide privately, and only the final execution is recorded.

In the example below, Joel submits an RFQ for 1 ETH. The market maker scans venues, including AMMs, CEXs, and other dealers, to reduce price risk. He then quotes the price of $3000 to Joel. If Joel accepts, the market maker hedges his position by buying ETH at a lower price, say $2,999 (not guaranteed).

The anticipated spread is $1. The quote is live for 30 seconds. If Joel accepts, the transaction is sent on-chain and settled atomically. Note that the market maker can only hedge once Joel signs the intent.

When the order flow consists of legitimate buyers, these takers receive competitive execution. Makers earn a small but fair spread. Makers price tightly because they’re confident they won’t be picked off and can keep the quote live for longer.

How Bebop Differs

Bebop tackles toxic flow through several defensive layers. The first is the last look, borrowed from traditional finance. When a user requests a quote, they provide a signature committing to the trade. The market maker then provides their own signature to complete it. But crucially, the maker retains the option to reject the trade if the market has moved adversely between quote and execution.

This disproportionately affects arbitrageurs who only submit trades when prices have shifted in their favour, while having minimal impact on regular users executing near the quoted price.

This dual approach matters because different types of trades benefit from different liquidity sources. A large swap into a major token might get better pricing from market makers who can internalise the order and manage their inventory efficiently. By maintaining both channels, Bebop captures the advantages of each.

So, where will you trade in the future?

The trajectory seems clear. Pure AMMs, as they are, face an existential version of the lemons problem. As markets mature with more sophisticated participants, the toxic flow will continue to drive good liquidity away from passive pools. We have already seen this play out. The most significant trades, the most professional users, and institutions tiptoeing into DeFi are not using raw AMM pools. They are routing through aggregators that tap into RFQ liquidity alongside automated market makers.

This is where Ben Thompson’s aggregation theory begins to show itself. It says that when the supply is cheap to spin up, value flows to the entity that gathers this supply and maintains a relationship with the user. RFQ venues aggregate AMMs, market makers, centralised exchanges, and DEXs with order books. They see the most traffic, understand which makers are best suited for each type of order, and they serve as the default entry point for users. Once that loop tightens, it becomes difficult for a standalone pool to compete for serious size on its own.

Sure, for simple swaps, AMMs are fine. However, for anything significant, exotic, or during volatile periods when prices fluctuate significantly before blocks are finalised, we’ll want an RFQ. The future looks like aggregation layers that understand this distinction, routing each trade to the location where it can be executed most efficiently. Obviously, Uniswap will still exist, but it will increasingly serve as one liquidity source among many, with RFQ market makers filling the gaps that formula-based pricing cannot handle.

I see no reason why an RFQ should stop at spot markets. Consider options for a moment. Sophisticated derivative instruments are used tremendously in traditional finance. Barely has PMF in DeFi. Options pricing requires sophisticated models that account for volatility surfaces, time decay, Greeks, correlation structures across strikes and expiries. You cannot encode this into a simple bonding curve. Market makers who understand these dynamics can quote options via RFQ, bringing genuine two-sided markets to on-chain derivatives.

The same logic applies to structured products, anything combining multiple instruments into a single payoff. These products are too complex for AMMs, but they are perfect for RFQs, where a market maker can price the entire structure and offer a single quote.

Tokenised assets or RWAs are a consensus bet in DeFi. They will become practical only with an RFQ. Traditional finance hates locking up liquidity. Every dollar that is locked could be earning yield or doing something productive. So, it is doubtful that only AMMs will house RWA liquidity. RFQ lets you bring the pricing models from traditional markets on-chain, with market makers who understand both worlds providing the bridge.

What is emerging is a more mature market structure. DeFi’s early phase needed AMMs because it was in a stage where nobody believed in us. There were no institutions, no market makers, no professional LPs. But now, they are here. As markets mature, the infrastructure needs to get more sophisticated. RFQ does not replace AMMs so much as it completes the picture, filling in everything that passive formulas cannot.

We are creating a financial system that combines the transparency and composability of DeFi with the efficiency and sophistication of traditional markets. You could trade anything, of any size, with confidence in execution. Market makers compete on price rather than on who can extract the most MEV. Users receive genuine best execution, rather than hoping their transaction does not get sandwiched. And new markets become possible that serve needs we have not even articulated yet.

This is how DeFi evolves without sacrificing what made it interesting in the first place. The infrastructure becomes more sophisticated, but the fundamental openness remains. Anyone can still participate. The protocols stay neutral. The difference is that now there are real professionals providing liquidity, not just passive pools hoping they do not get picked off. The lemons get filtered out, and what remains is a market where quality can survive.

Learning how payments scaled on the internet,
Saurabh Deshpande

Hello,

In the three years that we have been writing this newsletter, we have never encountered a moment where so many disruptive factors converged simultaneously.

On one hand, everything is being financialised. Pokemon cards? There’s a market for it. There’s one for betting on where Bitcoin’s price will be 15 minutes from now. There’s even one to bet on whether Jesus Christ will return this year. On the other, we are seeing a rapid decline in the accrual of value to tokens. The marginal bids from retail participants on exchanges are now with equities, ETFs, and meme assets. This shift changes how we as an industry think of business models and value capture.

We are also witnessing the rise of AI-induced slop taking over feeds at a pace like never before. When the cost of expression is low and the medium for doing so is easily game-able, everything said becomes a (growth) hack. Open your X feeds, and you’ll see countless views on how anything should be done. The means to produce noise have evolved, but the ability to filter signals has been on the decline. Our brains were never equipped to consume so much and understand so little.

Human attention has never been under attack more than it is now.

How does this translate to our work? Internally, we have identified three seismic shifts:

1. We will be tinkering with briefer articles with dense insights where the message is better served in a short form. Not everything has to be a 40-minute long read. Early on, it made sense to invest in long forms as it helped differentiate. We now have a critical mass of audience that allows us to tinker with briefer, quick snippets from founders.
   

2. We will be more explicit with our commercial investments. We have been committed to being the best partners to our portfolio of companies as narrative builders and strategic partners. Now that we are moving towards capital allocation, you will see more of our investments being covered on the newsletter. That involves communicating with our audience about where we deploy our capital and why. It is a natural evolution.
   

3. There will be a renewed focus on community. Our Telegram community is now close to ±5000 members strong. Our venture network is close to 100 firms. In this age of AI slop, communities that capture the attention of real humans, with reputations tied to them, will continue to stand out. Our in person events in 2025 were an effective way for us to meet and learn from many of you. Expect more of that. More real humans. Real. And human.

With that out of the way, I wanted to share what the focus areas for us internally look like over the coming year. In 2025, we mostly focused on the emergence of revenue in crypto and its economic ripple effects within the industry. In 2026, that equation inverts.

We believe that the crypto industry has reached a point of maturity where it is able to produce tangible impact on the broader economy. Firms that are able to integrate crypto into its operational flows will scale faster or see operational efficiencies that improvise unit economics. We see this playing out in three forms

The Stable-coinisation of Everything

Blockchains are structurally money movement technologies. Stablecoins are the most preferred form of moving money with this technology. With an entrenched duopoly of Tether and Circle, the sector is likely to start seeping into conventional business models requiring capital flow. The lowest hanging fruit here is micro-incentivising individuals with stablecoins. EarnOS rewards users in stablecoins for engaging with brands. The stables can then be used directly through Apple Pay.

Similarly, Architect Exchange allows traders to build exposure to currency markets (like USDEUR) and settle in stablecoins. Such models of value transfer cut down on conventional pay-out periods for traders and consumers alike. We anticipate more web2-esque businesses that benefit from faster money movement through stablecoin integration to emerge this year. In emerging markets, an inbound wire can take up to 60 days to clear. A stablecoin transfer takes two minutes on Ethereum and even lesser on other EVM and non-EVM chains. That is tangible progress. That is time the money could be reinvested into more productive use.

Enterprises Using Blockchains

Much like the internet, blockchains will go through phases of bundling and unbundling. In the last two years, everything had to be bundled onto singular networks as that is where (trader) liquidity lay. Capital formation through meme-like assets on Solana or Base determined where users went. In the next few years, we will see a period of unbundling where enterprise blockchains see more inbound liquidity. It is not so much about enterprise blockchains, but more about enterprises using blockchains to expedite functions that took forever.

Zk Sync’s Prividium allows enterprises to have customisable networks deployed with privacy. The Depository Trust Company is now moving towards tokenising assets on-chain. Nasdaq joined Canton recently as a validator. Each of these are moves that indicate that more traditional, institutional allocators will use blockchains albeit with trade offs. For founders, the choice is between the permissionless of a public blockchain or building the future of fintech with an extremely regulated partner that sees blockchains merely as infrastructure. Intercontinental Exchange (ICE) backing Polymarket is an extension of this phenomenon

The focus on revenue in 2025 was a precursor to where crypto is headed. DATs were a way for large capital allocators (like venture funds) to access public market liquidity. For founders, the mechanism of doing the same would be to partner with enterprise blockchains, and avoiding tokenisation altogether. Instead, their margins will come from the take rates on individual products that are sold through institutional outlets.

Will this be the antithesis of open money? Yes but it is also a direction towards scale. Ultimately, many talented founders may lean towards enterprise blockchain ecosystems in the coming years as scale is more valuable for them than idealism.

New Market Primitives

Up until last year what could be done on-chain was throttled by limits of scale. Transactions were limited and fees would escalate if the number of users on a product surged past a point. This is not taking into account the lack of on-ramps and clarity around stablecoins in multiple jurisdictions. But with the Genius Act passed we might see more unique instruments coming on-chain.

This will play out in two forms. First is niche-specific financialisation. USDAI enables the issuance of loans against GPUs. Glow uses game theory to incentivise solar farms. We will see founders with deep knowledge of individual sectors leveraging blockchains as a mechanism to raise capital and deploy them in productive ways that were previously not done.

The second is geo-specific financialisation. Aspora - powers 4% of global remittance volume with a focus on 700k Non-Residential Indians. Understanding cultural nuances and regulatory structures will be key for such businesses to scale. They could expand into being a neo-bank that plugs into DeFi markets for yield and offers commodities, currencies, and equities by using perpetuals (perps) as a settlement market.

This also explains why we have been looking at perpetual exchanges. One of the means through which non-conventional financial products like energy markets could find avenues to hedge, trade or settle on would be perpetual exchanges. Currently, players like Hyperliquid and Lighter are focused on commodities and currencies. The extension of asset types to new ones that are not readily accessible in traditional markets is what will help them scale.

The next phase for crypto will involve more boring things that scale exponentially while building moats.

In many ways, this part of crypto rhymes a lot with the internet. But unlike what most VCs suggest ,we don’t think it’s closer to 1998. We think it is 2004. Firms have bundled and begun tapping into public market liquidity through IPOs. More institutional allocators are looking at the industry whilst demanding clear growth metrics. Early entrants have either booked profits and left, or lost it at the casino. New ones want more tangible measures of value generated.

Every industry has a phase where it goes from pursuing idealism solely to economic value creation because capitalism demands that sooner or later - we generate value that is worth paying for.

Crypto is currently undergoing that transition. From a cluster driven by idealism to one earning its keep amongst technologies that have matured enough to matter.

As for us, what it means is we have begun filtering ideas and our pursuits a bit more harshly. The tolerance for hypothetical business models have now been replaced with curiosity for what makes systems scale. The real and tangible. As storytellers, you can live in the hypotheticals. As capital allocators, you cannot. There is a thin line between conviction and blindness. If the market is an objective measure of truth, you want the stories you tell to be as close to the truth as possible to be able to weave both storytelling and capital allocation properly.

In some sense, that sums up where our focus lies for the year.
On finding and sharing more truths about systems that scale efficiently.

We will be back to regular programming on Monday with a breakdown of how RFQ systems work..

Sliding into the weekend,
Siddharth Jain

Hey there,

This time of the year is like Thursday every week. Everybody knows the weekend is just around the corner. Some have dates planned. Some look forward to meeting family. And many simply wonder what to do with the existential dread that comes with holidays. Writing over the next few weeks will be choosing to battle the conflicting demands of attention from family and brands putting out reports.

So we will be taking a break until early January to plug off. There are birds outside the office we need to catch up with—the mountains beckon. Breaks and plugging off are crucial to the creative process.

As has been the tradition for the past two years, here is a PDF of everything we wrote in 2025 for those who want to export it to a reader, print it, or feed it into an LLM. If you are new here, you can grab the 2024 issue here and the 2023 version here. You can get the epub version and PDF version on Google Drive here in case you want to sync it to an external reader.

Decentralised

67.9MB ∙ PDF file

Download

Download

This year, we published fewer things, but with higher accuracy. For instance, we wrote about interface fees and embedded markets in April, months before prediction markets took off. Saurabh wrote about mergers and acquisitions in May, a quarter before Coinbase decided it was the season to buy out startups. In August, we wrote about the emergence of neo-banks. We wrote about the divergence of protocols generating revenue and ones without in July, almost two quarters before it was trending on X.

Admittedly, we may have been too late to catch the DAT party.

The Wall Street-ification of crypto is a trend we are watching from a distance and building a better understanding of.

Everything we wrote boils down to three themes that represent the disruptive cycles crypto is undergoing. These themes guide how we think about startups and our involvement with them:

Distribution and Scale Are King
When a technology first emerges, simply using it makes it novel. Mobile apps and the internet were once novel. So were crypto apps. But those days are behind us. Firms and founders are now judged by the scale and moats of their distribution. They are also heavily punished if they do not have it. The reason why many tokens are in down-only paradigms post-listing is that they confuse exchange listings with distribution.

Margins Matter
The most common business model in crypto today is interface fees on trading products. Farcaster had to pivot from a social network protocol to a trading-oriented wallet. Hyperliquid is currently in a shaky position with Lighter competing on a zero-fee model. All apps will undergo a period of creative destruction as they compete on lower margins to acquire and retain users.

For founders, the most complex problem will not be scaling in markets with multiple peers. It will be creating margins that justify their valuations. We will see the emergence of two kinds of crypto in pursuit of margin. One would be cypherpunk money, which commands a premium due to the extent of decentralisation and privacy it offers. The other is what we call - enterprise cash. These would be real-world assets and fintechs being built with a blockchain in the back-end capturing small slices of high transaction volumes.

The Niche Leverage
Where does margin come from then? It will often be due to vertical integration. USD.AI focuses on equipment financing, with an emphasis on GPUs. Qiro underwrites lending in private credit. Semiliquid is building infrastructure for lending between institutions in crypto. As the infrastructure layer matures, founders will focus on niche pockets for higher margins and reduced competition. As Thiel puts it - competition is for losers.

So, what’s in plan for 2026?

There are a few broad strokes that affect storytelling in crypto today. For the first time, our niche industry is eating into fintech. And with it, the audience base we can write for is expanding. The way stories are communicated is no longer just for the person buying tokens on Binance. We are writing for everyone who takes an interest in how the way money moves affects systems. The market has expanded faster than our minds could grasp that shift.

Crypto has always been a global phenomenon. We are writing for the person in Mumbai as much as for the person in Manhattan. We want to continue to make the newsletter a breeding ground for great ideas, good people, and the leverage of capital to come together. We want to be the hub, where writing is still human, bleeding with curiosity - and probably riddled with typos in the age of LLM-generated perfection. We want to speak to a global culture, as much as we are constantly defined by it.

With that in mind, here are two core priorities for 2026:

A Galapagos for Antimemetics

Do you want to be famous, or be understood? What matters more if you are a startup? In the age of LLM-generated content and declining attention spans, long, well-researched, humane stories will continue to have their place. As our feeds become hostile ground, the written medium will serve as a wedge for uncomfortable conversations, sharing unpopular ideas, and fostering collaboration—a corner for nerds in the age of glitz and glam.

Nadia Aspraouhova defines these forms of ideas that are often not viral as antimemetics. Memes and antimemes are the yin and yang of the attention economy. We want to continue making the newsletter the best place for long-form research and discussion of unpopular, often uncomfortable ideas. As crypto goes mainstream, the publication will remain a curated, safe corner for people to enter and explore the industry. We may lean into the memetic side of distribution on X going forward because who doesn’t like attention? Distribution, after all, is influence.

Liquid Markets Investing

One of the ideas we teased in Crypto, capital and culture, is that a culture is preserved or represented best when it has capital backing it. As writers, our interests are best protected when there are healthy margins in what we spend time on. This is where our interests come into conflict. Or what the big kids call - a conflict of interest.

If we believe in what we publish and the accuracy of our research, we should be putting capital behind it. And that is what we are doing. As we pen these words, we are in the last leg of signing one of the largest term-sheets we have invested in from our entity.

In 2026, you will see us speaking more of the bets we have made and the ways we get involved with our portfolio of companies. More growth stage, less pre-seed - in line with crypto’s current expansionary arc. Capital and effective storytelling are two levers that help shape reality. As DCo grows, we want the two to intertwine.

Our priorities don’t exactly change year to year, as much of the work we set out to do remains to be done.

Most people don’t realise the extent to which crypto has grown in the past three years. In May 2022, we saw $40 billion in value wiped out because of a stablecoin that rhymed with a Ponzi scheme. In November of that year, the largest exchange disappeared, taking user deposits with it. As late as December 2023, the most remarkable thing on Solana was a dog with a hat.

In contrast, under the current administration in the United States, fintech firms have embraced stablecoins, Robinhood is integrating prediction markets, heck - ICE invested in Polymarket, Arbitrum is bringing thousands of stocks on-chain, and the DTCC is working towards settling treasuries on Canton. We are witnessing what can be considered cultural inertia. That feeling of not wanting to give up a way of being, during the early innings of a sector coming of age.

Crossing the chasm is as much a psychological feat as it is a technological and socio-economic one.

I don’t mean to imply that in 2026, we will have to swap our NFT display pictures on Twitter for suits. But we do need to realise that the ways of building a business, capturing value, and scaling it have shifted. For us at DCo, the priority is to stay slightly ahead of the curve as it evolves, so we can share what we see with our readers and apply it to our portfolio of companies that has now grown to over 25 in size. Perhaps that is the cost of winning.

In a year marked by change, we hope you found the fortitude to evolve. If these changes have disrupted your life, we hope you find the strength and calm needed to regain stability. In 2026, we hope to continue to be inspired by beauty and guided by the curiosity that has marked much of our writing. We will see you in January - with long form articles, more depth, more quadrants, and perhaps messier art accompanied by a bundle of typos.

Signing off for now,
Joel, Sid - and the DCo Team

Before we begin, a big shout-out to Oliver Jaros from CMT Digital, who’s our guest author for this edition.

A few weeks ago, we wrote about how protocols like x402 are enabling machines to pay each other. Trust is a requirement for any payment network to scale. Visa allows refunds if fraud occurs. There is no “contact the manager” button on Metamask when a hack occurs on Ethereum. How do we build trust in a system dominated by agents paying one another? Between the previous piece and this one, there’s a common theme. We believe that stablecoin adoption will skyrocket in the agentic economy.

For this article, Oliver and I set out to explore challenges for stablecoin payments in the age of agents. We lay out a mental model for the wedges startups can use to crack scale in the agentic economy.

At DCo, we are seeing early consumer-oriented experiments in this theme. Drop us a DM on X - if you are building at the intersection of payments, crypto, and AI, and want to chat.

Let’s dive in!

In Pricing the Internet, we argued that machines will pay when metering is frictionless. Humans did not fully embrace micropayments due to the hassle and mental capacity required to pay attention to metering. But machines are different. All they see is 1s and 0s. Mental capacity or context switching doesn’t affect their ability to execute tasks. If going to sub-cent granularity makes their processes more efficient, they will do it, unlike humans.

We ended the previous article with a question: What happens when the agent screws up? It doesn’t matter whether the agent’s intention was right. At the end of the day, we are not supervising agents every step of the way.

We are left in a gap where new tech misses out on the good part of the old infrastructure: the ability to undo payments when things go wrong. This is what this article is about. We discuss what is needed for agents to be autonomous, who is building the infrastructure for it, and why new startups will emerge at the intersection of blockchain rails and autonomous agents.

Emerging Standards

Any form of commerce involves three parties – buyer, seller, and the middlemen who facilitate their interaction. Middlemen can be platforms or marketplaces like Amazon, or card networks like Visa that facilitate payments.

Buyer Side

Consumer applications tend to handle the money or transactions and get a cut in the process. But what happens when the consumer is AI acting on our behalf? There are a few emerging standards that are searching for the answer.

ChatGPT has 700 million active users trying to get something out of AI. While we don’t buy or sell things through an agent interface yet, but use it for discovery. Whether it’s buying running shoes or looking for hotels in El Calafate, I’ve been using AI to compare things. Having the option to buy things from the same interface makes it way more convenient. And that’s what OpenAI is doing with the Agentic Commerce Protocol (ACP) in partnership with Stripe.

This is the most straightforward way agents are handling money today. The user remains in control every step of the way. When someone places an order, ChatGPT sends the necessary details to the merchant’s backend using ACP. The merchant then accepts or declines the order, processes the payment via their existing provider, and handles fulfilment and customer support just as they do now.

Think of commerce through ACP as allowing your intern to spend fixed dollars, with you authorising product or service selection, the merchant, and the final payment.

While OpenAI and Stripe have ACP, Google offers Agents to Payments (AP2). Before we delve into that, let’s zoom out a bit. What Google wants to address is the interoperability problem. Currently, AI agents operate in silos. Gemini doesn’t talk to Claud. ChatGPT has no clue what’s going on in Perplexity. You get the point.

Ideally, we want these agents to communicate in a common language when tasks become complex, and we need them to work together. Google developed A2A (Agent2Agent), a protocol that allows different agents to communicate and coordinate.

But it’s not enough for agents to be able to talk to each other. They need to be able to use tools, access APIs, and services. Model Context Protocol (MCP) allows agents to use tools like Google Calendar, Notion, and Figma.

MCP defines a common language. Agents can use any tool without any custom code as long as they all speak MCP. Anthropic built it, but the specification is open, and companies are adopting it quickly. An MCP server is essentially a translation layer that sits in front of a company’s existing API. It exposes the company’s services in a standardised format that any MCP-compatible agent can understand.

Okay, let’s get back to AP2. A simpler way to think about what it enables is that MCP gave agents the means to access data, files, and tools. A2A gave them a voice. They could communicate with other agents. AP2 gives them a wallet. They can now spend money securely.

All these protocols keep the user in charge, with agents having limited spending power. This solves distribution and flow. But neither one has figured out what happens when the agent gets it wrong.

Seller side

Buyer’s side is only half the story. Different standards are emerging on the seller side. These focus on how machines pay for access to APIs, data, and content.

The most talked about right now is the x402 standard, an open protocol that Coinbase developed. It revives HTTP status code 402, which was defined way back in 1997 as “Payment Required”. But, it was never actually used. x402 brings it to life by pairing it with stablecoin payments that can actually settle micropayments economically.

x402 turns an HTTP request into a paid request. Whenever payment is required, the server asks for it. Since the agent has a pre-allocated budget, it pays the server in the same flow and receives the data. This makes pay-per-request or pay-per-call practical for machine commerce.

With x402, an agent can pay for exactly what it needs in that moment. For example, a single article behind a paywall for two cents. One API call for a fraction of a penny. The transaction settles in seconds on-chain with no ongoing relationship required.

Cloudflare took this concept and built something more specific with their Pay-per-Crawl system. It uses HTTP 402 under the hood, but it matters because of Cloudflare’s dominant position in the market. 20% of all web traffic goes through Cloudflare, which gives them enormous leverage.

Pay-per-Crawl uses Cloudflare’s edge to require payment from AI crawlers before serving the content. This turns access into enforced metering. Publishers have been watching their traffic crater as people stop clicking through from search engines and just read AI-generated summaries instead. With Pay-per-Crawl, they can charge the AI labs directly every time a crawler visits.

Card networks are also trying to extend their existing rails to handle agent transactions. Visa announced an MCP Server plus an Acceptance Agent Toolkit. Mastercard has a program called Agent Pay. Both are extremely early, still in pilot stages, but they matter because Visa and Mastercard already have global distribution, issuer relationships, and merchant acceptance everywhere. The basic idea is to register agents, add spending controls, and enable agent-initiated transactions on the same card rails that handle human purchases today.

Urgent Need to Fill the Trust Gap

All of these standards assume that the payment will go through smoothly and that the outcome will meet expectations. ACP and AP2 keep humans in the loop during checkout, providing some safety. x402 variants handle machine-to-machine data access, where the stakes are typically lower. Card networks extend their familiar protections, but they come with slower settlement times and higher fees.

Speed is the primary goal when enabling micropayments at scale. Card rails take days to settle and cost merchants a few percentage points per transaction. Crypto rails settle in seconds and cost fractions of a cent. But this efficiency comes with irreversibility. Once a crypto payment is completed, there’s no going back.

Commerce has an entire infrastructure built around the possibility of mistakes. When you buy something with a credit card and it goes wrong, there’s a process. You call your bank. The bank disputes the charge. The card network investigates, holds the funds in the meantime, and either refunds you or sides with the merchant. 261 million transactions were disputed with a total value of $34 billion in 2025.

Agents operating on stablecoin rails have none of this.

This becomes more complicated when agents start coordinating with other agents. When hundreds or thousands of multi-agent workflows are involved, untangling liability can become a nightmare.

Card networks won’t take on that risk, at least not at their current economics. Visa and Mastercard’s agent programs still carry the standard interchange fees and multi-day settlement times. They could pivot to instant stablecoin settlement, but doing so would mean giving up the dispute infrastructure that justifies their fees.

The traditional financial system didn’t come with built-in dispute resolution mechanisms. Diner’s Club, the first credit card, was issued circa 1950. But it took another 24 years before consumers could dispute transactions. The modern infrastructure we take for granted today emerged gradually as problems surfaced and institutions responded.

Agentic commerce doesn’t have that luxury of time. API requests already account for 60% of the dynamic HTTP traffic that Cloudflare processes. Bots and automation make up nearly half of all web traffic. ChatGPT’s 700 million users can now check out directly through ACP with Etsy, and Shopify integration is coming. The volume is already here. There is latent behaviour to use agents for tasks. It won’t be too long before agents are used for commerce.

So we can let the legacy financial infrastructure continue with longer settlement times or deliberately build trust infrastructure that complements quick blockchain settlements. The former will keep agents from realising their potential. The latter is an opportunity and a logical extension of how agentic commerce will progress.

So what does that actually look like?

Not surprisingly, there are two parts to this. Before and after the transaction.

Before the transaction

Whether an agent should even be allowed to transact depends on three factors: identifying the counterparty, employing fraud detection, and using reputation scores to make informed choices about pricing and access.

Plaid connects to one in two US bank accounts and processes a million account connections every day. When you want to prove to Venmo who you are, it uses Plaid to verify your identity.

Currently, any agent that interacts with an API or attempts to scrape the web or make a payment has no equivalent. All the server sees is some vague ID, like a wallet address or an API key. It has no idea who is making the call. There’s no consistent identity that could build up a good reputation across different services, so every new interaction starts from scratch with zero trust.

American adults have lost approximately $47 billion to identity fraud in 2024.

A Know Your Agent (KYA) layer would work similarly to how Plaid became identity infrastructure for fintech. It would issue persistent, revocable credentials that link an agent to a human or organisation.

Card networks have spent decades building systems that can spot suspicious patterns across millions of transactions. They know what normal human spending looks like and can flag anomalies in real time. If an agent gets compromised and starts making unauthorised purchases across multiple merchants, there’s no shared fraud graph to catch it.

Visa says it blocked $40 billion in attempted fraud after investing $11 billion in securing its systems over 2019-24. Stripe processes over $1.4 trillion in payments annually and trains its Radar fraud detection system on that entire volume. During Black Friday and Cyber Monday 2024, Radar blocked 20.9 million fraudulent transactions worth $917 million.

The fraud detection layer doesn’t exist for agents. Whenever an agent is making an x402 payment, no shared fraud graph flags suspicious behaviour, such as higher-than-normal spend or higher spend frequency.

Without a persistent identity and reputation, every agent interaction starts from zero. Reputation is deeply embedded in human commerce. You get ads based on the reels you scroll. Your Uber star rating is a factor when drivers accept your request. The credit score you generate by borrowing money and paying off loans follows you to every financial institution. It shouldn’t be any different for agents.

After the transaction

Chargebacks are how card rails handle disputes by reversing funds from the merchant after a customer disputes a credit/debit card transaction with their bank. They get abused, too. Chargebacks cost merchants an estimated $117.47 billion in 2023. For every $1 lost to chargebacks, businesses typically incur at least $3.75–$4.61 in total costs, when you factor in fees, lost merchandise, and administrative overhead.

Merchants win only 8.1% of the disputes they contest. 84% of customers find filing chargebacks simpler than requesting refunds from merchants directly.

Stablecoin transactions initiated by agents will settle in seconds and there is no recourse right now. Cloudflare has proposed deferred settlement extensions to x402, which would allow for hold periods before funds move irreversibly.

Developers are already building these infrastructure primitives. At ETHGlobal Buenos Aires, a team built Private-Escrow x402. The escrow scheme has buyers deposit funds into a smart contract upfront and sign an intent off-chain at the time of payment. A facilitator batches hundreds of these signed intents into a single settlement transaction, cutting gas costs by a factor of 28.

But that’s just an infrastructure primitive. Someone needs to productise it.

Who builds all of this?

I can’t help but think of a time when carriers dominated the telecom industry. They had billing relationships with every phone owner, and yet they missed generating value from smartphones. They captured distribution and mobile advertising. Together, they generate hundreds of billions in revenue that telecom carriers could have captured.

Card networks face a version of this problem now. Visa and Mastercard have spent decades building exactly the trust infrastructure that agentic commerce lacks. But their entire model depends on interchange fees that only exist because they control the payment rail. They spend billions maintaining and improving this infrastructure, which can only be funded by a few per cent transaction fees. Building consumer protection for stablecoin transactions would mean subsidising a competing rail with none of the revenue attached.

If the card networks are not going to do it, the next incumbents are AI labs like OpenAI, Google, and Anthropic. All these labs want their agents to be used everywhere, with every merchant and across every protocol. Running a centralised identity registry would mean they become liable when those agents misbehave, and they do not want to become the court of record for your booking mistake.

They would much rather someone else build identity and recourse as infrastructure they can plug into, the same way they plug into payments or search today.

Cloudflare occupies a strange position. They already see a massive share of web traffic. They already run bot detection. Their AI Audit tool lets publishers track which crawlers access their content and how often. Extending from “identify the bot” to “verify the agent’s credentials and reputation” is not a huge technical leap.

But Cloudflare has always positioned itself as neutral infrastructure. The moment they start issuing trust scores or adjudicating disputes, they become something more like a regulator. That’s a different business with different liabilities.

Startups exploiting the three wedges

You won’t beat OpenAI with model quality or Cloudflare with more traffic. You will have to find something in the stack that their business model doesn’t allow them to touch (at least for now) and still has value. I think those wedges are – identity, recourse, and attribution.

Agent identity is the most straightforward. The registry model is proven. Although Plaid is a cliché example, it fits well. They did it for bank accounts. A startup could do it for agents. Issue credentials, let them accumulate reputation, and give merchants a score to check before accepting payment. The defensibility comes from network effects. Once enough merchants check credentials against your registry, agents will have no choice but to maintain good standing.

Recourse is harder because it requires taking on risk. Think of it like insurance. You collect a small fee on every transaction and absorb losses when something goes wrong. Volume is the name of the game. Card interchange runs 1.5% to 3% and covers dispute resolution. Stablecoin rails cost a fraction of that, so a recourse layer could offer comparable protection at 50 basis points and still have margin.

Attribution is the most speculative of all, but it will eventually be built. When agents start influencing purchases, brands will pay to shape what gets recommended. The auction mechanics are designable. But this one has a cold start problem, the other two don’t. You need brands, agents, and merchants all participating before the market functions.

These wedges matter differently depending on where the agent economy sits. Identity is critical when agents act without human approval for every transaction. Recourse matters whenever agents start moving real money. Attribution only kicks in when agent-to-agent commerce reaches enough volume to support an advertising market. This brings us to what the trajectory actually looks like.

Startups will build parts of the agentic economy infra

The journey of agents can be mapped in three phases.

1. They first act as interfaces.

2. Then they execute under human supervision.

3. They transact with each other autonomously.

We are in the first phase. ChatGPT’s Etsy checkout integration is a good example. We browse products through conversation in a chat interface (though not everyone is doing that yet), the agent surfaces options, but we humans pull the final trigger. Trust is borrowed entirely from existing infrastructure.

This phase belongs to incumbents because it’s a distribution game. The value accrues to whoever owns the interface where purchase decisions happen.

The second phase is marked by agents getting more autonomy. Instead of only suggesting the itinerary, agents book flights, car rentals, and hotels. We provide intents or constraints, the agents execute, and we verify the final results.

This is where the trust layer becomes essential. Users will not delegate authority to agents without recourse. Merchants will not accept agent-initiated payments without some way to verify the agent’s identity and authorisation.

This is also where startups have a real opportunity. Incumbents perhaps don’t have the clear incentives to direct their resources towards building trust infrastructure for stablecoin rails because tremendous growth lies ahead of them within the existing phase, where they dominate.

OpenAI generated $13 billion in revenue this year. For context, Tether is on track to make more than that in profit, with $10 billion in profits until October 2025. The identity, recourse, and attribution layers will emerge from new companies solving specific problems at the boundary between what agents can do and what users will permit.

The third phase is autonomous agent commerce. Your agent doesn’t ask permission for routine decisions. It negotiates with other agents, bids for compute and bandwidth, participates in ad auctions, and settles thousands of small transactions continuously. Stablecoins fit the bill for the default settlement layer because nothing else handles the volume, speed, and granularity that machine-to-machine commerce requires.

In this phase, the differentiator is no longer who has the best model or the fastest chain. What matters is who built the most trusted infrastructure. The passports that agents carry. The courts that adjudicate disputes. The credit systems allow agents to transact beyond their immediate balance. These institutions for software will determine which agents can participate in the economy and on what terms.

We’ve built the plumbing for agents to spend money before building any way to verify they should be allowed to. HTTP 402 sat dormant for thirty years, waiting for micropayments to make sense. Now it works. But the trust infrastructure that enables human commerce, including identity verification, fraud detection, and dispute resolution, lacks an agent equivalent. We solved the easy problem first. It will take some time before agents are ready to do business with each other.

Looking forward to hiking during the holidays,
Saurabh Deshpande