Should protocols make money?
Sid and I discussed the Fat Protocol Thesis a few months back. Much has been written about its validity. For those not in the know, the thesis argues that the value of a protocol will be higher than the combined value of all the applications on top of it. It is an excellent narrative to buy into when a new network is coming to market.
Joel Monegro wrote it in 2016 when ETH's price was a few dollars. There has been much debate around the thesis since then, but if I had to summarize, here's what happened. During bear markets, as value accrues to stable, less volatile assets, it is possible that applications atop a protocol become "worth" more than the protocol itself. A simple place to check this is the value of stablecoins held on top of certain L2s.
During bull markets, given that the underlying infrastructure (an L1 or L2) is an indexed bet on the applications built atop it, money flows towards investments in the infrastructure itself. It drives the underlying asset's price (like ETH or OP) and proves the thesis right.
We figured it was worth revisiting the thesis. Don't worry; we will not be arguing for or against it. A lot of intelligent people have done so over the years. I think it makes little sense to commit to that concept forever. It is called a thesis and not a law for a reason. Today’s piece explores “value”, how it is captured and why protocols may cancel out one another’s growth. It is a bit long, but there are memes along the way.
We’ll start with some basic definitions.
Protocols, Platforms & Applications
A protocol is a set of rules followed by participants within a system. For instance, a protocol in the military dictates how people should behave. Diplomats have a "protocol" for how they interact with one another. Think of protocols as a bundle of rules. In the context of machines, specifically computers, protocols have been defined as the rules that specify how data flows. RSS, for instance, is a protocol that defines how information about articles is updated in your client. SMTP defines how emails flow through to your inboxes. You get the gist. Protocols are bundles of rules that are context-specific.
On the other hand, platforms are operating systems, social networks (like Meta) or hardware (ARM/NVIDIA) that enables a group of protocols to run atop it. When you use Outlook (an application) on Windows, you use SMTP (the protocol) to get data to Windows (the platform). There are no Web3 native platforms that have scaled yet. Solana’s mobile devices may have their operating system, which is fine-tweaked to the needs of the industry. Ronin has its app store that allows for the distribution of NFT-enabled games.
But when you think of the kind of scale Azure, Facebook, or iOS has, there are no platforms at scale in Web3. (Quite possibly, because we don’t need them yet). There were multiple attempts to create hardware-enabled mobile devices by both Samsung and HTC in the 2019 cycle, but I assume that with the release of tools like Secure Enclave, the need for wallet hardware-enabled phone devices have declined.
What confused me was the concept of applications that can also be protocols. Take 0x, for instance. Is it a protocol or an application? Matcha is the application. And 0x is a protocol multiple DeFi products can tie into for liquidity. Similarly, OpenSea’ has Seaport, their protocol for NFT marketplaces of all kinds to tap into shared liquidity. Do you get the gist?
Because protocols do not find multiple developers on their own in the early days, developers often release an application with it to bootstrap activity. And there’s a reason for this. If you are a standalone application, you stand to be disrupted. OpenSea lost the royalty wars meaningfully to Blur. But if you are a protocol with multiple applications built atop it, the possibility of complete disruption reduces drastically.
So if you zoom out a bit, the playbook in the past few years has been relatively straightforward.
Release an application. Bootstrap liquidity by offering token incentives
Grow to a point where you can now allow third-party applications to tap into your liquidity.
Release a protocol which has a governance token.
Compound and Uniswap are both instances of this playbook. It just happens that the core products they released are so powerful that people don’t think of the applications built atop the protocol. Products like DeFiSaver, InstaDApp, MetaMask and Zapper send liquidity to these products. But the bulk of user activity happens on the initially released product—the native website of the protocols.
In these instances, the team builds moats in two ways.
First, through distribution, they have by being the most reputed brand in the industry.
Secondly, the network effects emerging from multiple applications sending them liquidity.
In other words, applications can evolve into protocols (or platforms) in the digital asset space. As roll-ups make it easier for applications to pretend to be L2s, we will see an increasing number of apps claiming to be protocols for the bump up in valuations they stand to see.
Utility Enters The Chat
During the ICO boom, there was no clear definition of what tokens would be doing. There was a general understanding that they should not be doing things that may turn them into a security, and that’s about it. People would tinker with dividends, buy-back and burn (like Binance) and governance rights that came with tokens. The crux of the problem was tying economic value with something minted for no cost.
Transactional networks like Bitcoin, Ethereum and Ripple can claim that a small amount of the asset is required for transactions. As the number of transactions rises, the base asset (ETH, XRP1etc.) would increase in value. This is a sound thesis if you have an exponential number of people trying to do transactions. It works because it costs the equivalent of a low-cost Android device to do a transaction on Ethereum if someone is busy minting kitties at the same time you are trying to get money across.
It was a helpful mental framing as many tokens were valued based on the revenue they created from transaction fees. For context, Ethereum’s EIP 1559 burns a small portion of the token’s supply on each transaction. Thereby, making it a deflationary network. This philosophy works exceptionally well when you are a base layer that derives value partly from the number of transactions you enable.
The challenge emerges when you are not a transactional network but an application. Requiring users to hold your native asset is a hindrance. Imagine if your bank required you to hold their stock each time you took a loan. Or if the server at McDonald’s asked you how much of their stock you owned before giving you a burger.
Tying an actual use case with a base asset leads to horrendous outcomes if it becomes a requirement. Exchanges understand this exceptionally well. It is the reason why Binance or FTX (RIP) never required you to hold their tokens for an exchange. They only nudged you towards it with a discount for using their tokens.
Many of what we now refer to as governance tokens are utility tokens in hiding. That is, their utility stems from the idea that they can be used to govern the network itself. Now it is up for debate whether true decentralisation of governance ever occurred in DeFi - but the base assumption is that holding an asset helps voice opinion regarding how a product is run.
For many DeFi projects, it meant being able to change fee variables, assets supported and other random functions involving the treasury. In such instances, a token holder receives no revenue from the product. But the token they hold “governs” a treasury which may get income. So if a product creates a hundred million in fees (for users), the argument is that a multiple based on that is relevant when considering a fair valuation. Compound and Aave's multiples align with what we see with listed FinTech companies. Markets are driven by narratives in the short run but return to rationality over time.
Markets are narrative machines that fuel up from time to time. When that happens, platform usage drives valuation less by the fees they generate, and more so by the narratives they can fuel. In simpler terms - a thousand people noticing ten people are using a dApp can drive the valuation of a token higher than the fees generated from those ten users can.
This is because, given the liquid nature of digital assets, there are more capital allocators than users. For a sense of scale, Compound has over 212k individuals holding their tokens in a wallet outside an exchange. In the last month, ~2k people used compound for a loan. That 1% ratio, is still a healthy number by Web3 standards.
Ashwath Damodaran refers to this syndrome as the big market delusion2. A paper written by him in 2019, explores how multiple VCs bet on a similar theme with the assumption that all of their bets will eventually become a winner. We are seeing this in AI these days.
Billions of dollars flow into multiple firms doing the same thing, assuming the market is large enough to sustain all of them. VCs pile on capital in hopes that the startup they have invested in will emerge winner, and have a large enough market share to justify higher valuations. Given the power laws we often see in ventures, many die. We see a variation of this with digital assets.
Individuals overcrowd into trading an asset on seeing a handful of users emerge. Often, the assumption is that utility will continue to rise and match up with where valuations are. Soon enough, a new product appears with a shiny token airdrop. Users flock elsewhere, and valuations tumble as the market reprices the lack of platform usage.
Sidenote: This paper from 2021 studied the ratio between speculative behaviour and platform usage for utility tokens. I won’t dive into it in this article as the paper was relatively outdated at the time of writing.
dApps vs Protocols
Now that we have established some baseline economics around how protocols and applications make money, it is worth looking at which of the two generates more in fees. The chart above excludes Bitcoin and ETH, given their early mover advantages. It also does not include Solana, in case you are wondering. You will notice that apps like Uniswap and OpenSea make considerably more in revenue than the average protocol. This may conflict with the idea that protocols should be valued more than applications, given how value flows downwards (to the infrastructure enabling it).
This is where blindly citing fat protocol thesis as a basis for backing new L2s becomes faulty. Mature applications on Ethereum can generate more fees than entire protocols of relatively younger age.
There’s a reason for this. dApps tend to make money by capturing a small percentage of the transactions on them. Your fees are proportional to the amount of capital flowing through the product and your take rate. Uniswap and OpenSea have made nearly $2.8 billion because they have a high monetary velocity (frequency at which assets change hands) and an enforceable take rate that passes on value to users.
In the case of protocols, enforcing higher take rates as usage increases breaks network effects unless the use case justifies it. Let me explain. It is acceptable to pay the equivalent of a week’s income in emerging markets for a transfer on Bitcoin if your life depended on it. Assuming users will be OK with that to play your Web3 game is faulty thinking. Bitcoin’s promise of immutability and decentralisation is a feature people would pay a hefty premium for.
The fee rate on Bitcoin is justified by
The lindy effect of the network
Its decentralisation and immutability.
But when you introduce something like a stablecoin issued by a centralised source, markets will be repricing what they are willing to pay on protocols. This is why Tron is a hub for stablecoin activity. Here’s one way to quantify it. The average USDC transfer on Ethereum last week was close to $60k. For those on Arbitrum, it was down to $9k. And on Polygon, $1.5k. One can debate using “average” as a metric here, but the assumption goes that transactions worth under $100 become possible as fees trend lower. The point I am trying to drive is
We presume protocols become more valuable as transaction count increases, and costs of transacting increase.
But the high costs tend to break the network effects of users concentrating on a single network and drive them elsewhere over time.
This is why dApps in emergent chains never see critical velocity to create sufficient fees. When you launch a DeFi product on Ethereum today, you are tapping into the network effects of users that have made their wealth from ETH, ICO boom, NFT boom and DeFi boom alongside the robust infrastructure that allows people to trade, lend and borrow. When you build on the hot new L2, you hope users bridge their assets and use your product. It is like building a business in a new nation-state. Sure, you have less competition, but there are also fewer users.
It’s like running the only Starbucks on Mars.
Profitable? Likely not.
Community as a Moat
We have been thinking extensively about what is a moat in Web3. Because, unlike other industries, most applications in crypto are known for two things.
Open-source: you are enabling anyone to replicate what you have built
Capital flight: you are allowing the users to leave with their money as soon as they please
Inspite of both these characteristics, Uniswap, Aave and Compound have maintained relative dominance in what they do over the years. Multiple DeFi products have forked Compound and failed miserably in the process. What is the moat for these products?
(Sidenote: Uniswap has begun using licenses that are restrictive in the past few months)
The simplest measure of a moat in the industry is liquidity. If you are a capital-intensive product, liquidity is the amount of money available to facilitate transactions on the product. If you are a consumer application, like a game, it is measured in attention. In both cases, what drives either liquidity of capital or engagement, is a community. The only real moat in Web3 is a community. And what keeps early community participants retained are capital incentives or product utility.
Exponentially better user experiences like ChatGPT do not require incentives for users to flock to it. Blockchain-enabled applications occasionally produce similar magic. DeFi crossed this chasm in the golden age of AMMs and permissionless lending, coming alive in June 2020. An era, we fondly remember as “DeFi Summer”.
A large user base looking to make a quick buck through an airdrop may look like a community. But it is not. It is a “cost” on the network in the long run because the buy-side for the token would have to give them sufficient liquidity if you want to maintain prices. For instance, yesterday, it was revealed that 93% of the tokens held in wallets for Arkham Intelligence moved their token right away. Are the members who sold - community members or a cost on the network?
They can be community members if they strategically buy back in. But as long as they don’t need the tokens to use the platform, they have no incentive to buy back in. They can allocate that money to hundreds of other tokens. DeFi products like Compound and Uniswap have a community not just in the form of token holders but also in the thousands of individuals that have kept billions of dollars in their product’s liquidity pools.
One can fork their code base, but you cannot replicate the liquidity pool, on a long-enough basis, without building a community with conviction. Capital incentives help retain communities in the long run.
Capital incentives can be tokens given in exchange for performing a function on the network. Those providing storage on Filecoin for instance, receive tokens for their contribution. Being early to networks is another way capital incentives accrue to users. Bitcoin and Ethereum are similar in this regard because their early adopters made abundant wealth by being early and having the patience to hold.
The need for capital allocation from users is transcended through shared culture. Bored Apes and the myriad of GMs or WAGMI statements we used to see on Twitter in the last bull run is an instance of this. Culture helps individuals align their identity, and retains individuals longer. There is no quantifiable way to measure culture, but the excitement one sees around EthCC or Solana’s Hackerhouses is an instance of this. It gives individuals a mechanism to build, relate and ideate without putting up capital to be a part of the conversation.
Protocols don’t run on vibes alone. You need people to build on them. Developers are how culture and capital combine to help develop tools that retain users in the long run. If you think of a protocol as a nation-state, developers build utility that retains users (citizens?) on the networks for the long run. Protocols can charge fees, much like highways can charge tolls. But if they become prohibitively expensive, they will drive users elsewhere. Looking through this lens, it becomes evident that protocols may not be designed to be making money at all in the first place if the use case is consumer related.
Moats in web3 emerge from users sticking to a network long enough to help facilitate economic transactions in the applications built on top of it. Every network has the same group of dApps replicated with the same code but different branding and sells the idea of “lower transaction fees” alone as a USP. We will soon have fragmented ecosystems with divided user attention. Indeed, capital will flow to it in the short run through exchanges, as users trade, but they will soon be dead towns like EOS.
In the real world, you cannot replicate a nation-state. No mechanism exists to expand the land area within national borders (without violence or economic alignment). This is why people are forced to concentrate in hubs, which have historically been ports. London, Mumbai, and Hong Kong - are all similar in this regard. This concentration of people helped drive network effects within the city. Rents surged, but it meant faster groceries and better services.
In the digital realm, users were forced to concentrate in an ecosystem due to a mix of how intellectual property rights work and expanding product suites. Google’s launch of their search engine, Gmail (2004), Android (2005), Youtube (2006) have each facilitated our stickiness to that ecosystem. A user signing up for GMail inevitably enters the rest of Alphabet’s product suite. Apple and Meta have similar strategies that concentrate users within their ecosystem.
Apple takes it to the extreme by owning the whole stack, from hardware to payments. Concentrating user bases allows the facilitation of economies of scale. There’s a reason why I mentioned this - and it boils down to developers.
If I drew a hierarchy of needs for early-stage protocols and dApps, it would look like the image below. You need developers to
Form the code,
Based on which capital can be invested into
To develop a lore
That attracts users.
Without the code, we are simply running in circles.
VCs that have been around since the early 2000s emphasise a lot on developers. It is because developer count is a distilled measure of the economic activities that may occur on a protocol. Say you purchased an iPhone for the camera. There’s a high chance you end up paying for an app that helps you edit images on the device.
So your initial decision (the camera) fuels a secondary purchase (the app). With each new app, the ecosystem grows stronger as the suite of products a user can use expands. The value proposition for buying the device is no longer the camera. It is the entire ecosystem that opens up to the user.
A variation of this was also evident during the early days of the web. People were not taking subscriptions to the “internet”. In their minds, they were getting a subscription to AOL - a group of landing pages summarising what is happening on the internet. But the open internet grew only when people realised they could send e-mails, send cringe texts to their crush from school and get bullied in games from the same network.
Now imagine if 20 different internets were competing, each with their listed equities and their variations of applications with entirely different branding and transaction fees. Consumers would be baffled, and the internet would not be what it is today. That is where we are at with Web3 native protocols.
Value Capture Another Day
Protocols need significant user liquidity to support emergent applications on top of them. Everyone is incentivised to pretend to be the next L2 in the age of roll-ups. But with each new protocol, we fragment the number of users web3 native products can have. There may come a period where users don't know the chain or stack behind their tools. But we are not there yet.
In the interim, wanting every protocol to capture as much fees as well-established dApps may be faulty. The first generation of blockchain dApps was capital-intensive. The next generation may be attention intensive. There are no Web3 games of scale, because we value transactions over great games. The same could be argued for on-chain media too. Blockchains are financial infrastructure by nature. So it is only fair to presume every user would want to transact.
But that mode of thinking is partly what's held the industry back.
All of this made me wonder what even is "value". Tokens, like equities, in-game items and any other liquid asset, will always have a premium. Depending on the crowd's hopes or fears, that premium may sway higher or lower. Speculation - has been a driver of financial markets for atleast eight centuries. And I doubt we will change this aspect of human behaviour soon.
For founders, the message is quite clear. There are two ways you can make money. One is through driving the narrative, even when there is a lack of protocol fees or usage. Meme tokens are a version of this, taken to the extreme. The other is building a dApp that generates fees and commands a reasonable multiple. Aave and Compound have matured to reach that arc. Both, require an incredible amount of work.
The best founders we have seen can drive both narratives and platform usage. Having only one is generally a recipe for disaster. Protocols or applications that offer an unrivalled core utility can have higher take rates due to their stickiness. This is where Peter Thiel’s view of “competition is for losers” come to play. The more crowded a market segment gets, the lower the probability of a new entrant being able to command higher take rates or sticky users. All of this looks at user concentration alone. What about protocol economics?
Joe Eagan from Anagram had a good analogy here. The best protocols have an incentive to behave a lot like Amazon. The marketplace made little to no money in net income for the longest time, but the inherent network effects paid off in the long run. The broadest range of sellers meets the largest group of buyers on Amazon. Any protocol that “succeeds” in the long run may have a similar attribute. Exceedingly low fees with the broadest range of applications built on top of it so that users don’t have to bridge elsewhere to complete their day-to-day functions. The monetisation could occur through the richness of the data such a protocol would leave on the blockchain. (That last bit can be debated quite a bit).
One way this could play out is with a protocol launching without a token. Instead of charging fees for a new native asset, it could charge in dollars. Imagine paying $0.0001 for a transaction, in USDC. Users “reload” their wallets with a dollar after every 10k transactions from a local store. The caveat is that most base chains cannot do this, as their native tokens are needed in their security model, and we don’t quite know how such a protocol would make money. Such a protocol could scale exponentially, without users having to fork over large sums of money each time protocol usage surges like you have with Ethereum today.
If we believe that blockchains are infrastructure for transactions, and all of the actions on the internet will become transactions, we may inevitably need low-cost protocols with fixed costs. Or, we may end up with 50 new L2s, each of which sees sky-high valuations, because markets like a shiny new thing. Markets can facilitate both. And that is perhaps the beauty of it. There is room for both utility and speculation.
I’ll see you guys next week with some work we are doing on algorithms and identity in the context of Web3 social networks.
Telegram, Pitch Decks & Referrals.
Join in with close to 5000 researchers, investors, founders & overall great human beings. We don’t exactly talk much, but it would help you stay close to what we are focusing on & connect with others building cool things.
Fill out the form below if you are a founder building cool things and in the process of raising money or looking for feedback on what you are pursuing. We like the builders.
Enjoyed reading this? Consider sharing with a friend for access to premium newsletters we enjoy reading and good karma.
I am not sure you can tie Ripple’s price with its usage even now. It is one of the great mysteries of life.
You might want to read this paper because I will cite it often in the future.