Written by: Decentralisedco
Compiled by: AididiaoJP, Foresight News
Today's story is about what makes protocols anti-fragile, and how tokens can be used as levers to drive ecosystems. In the weeks following the hacking incident, we delved deep into the Hyperliquid ecosystem. We quickly discovered that protocols in this industry are building formidable competitive moats through vertical integration.
This article explores how leaders in the Web3 space maintain their moats.
I frequently use terms like "supply-side aggregator," "demand-side aggregator," and "vertically integrated capital ecosystem" in this article. For clarity:
- A supply-side aggregator brings together differentiated market participants to offer commoditized products (e.g., Uber).
- A demand-side aggregator expands products to groups that appear similar externally (e.g., Amazon).
- A vertically integrated capital aggregator is a financial entity that integrates multiple parts of an ecosystem to provide users with various products in the same venue.
Blockchains are monetary rails. The value of a protocol lies in its economic output. Composability and real-time verifiability enable blockchain-native businesses to engage in vertical integration. Tokens allow participants across the stack to be incentivized through a shared medium. Teams that intentionally enable value capture in every part of the stack possess a moat. Vertical integration helps accelerate capital flow within the ecosystem. If done well, it becomes a source of revenue.
Uber aggregates passengers across the city, positioning itself as a demand aggregator, as does Swiggy. You might think there isn't much difference between an average passenger or a hungry person; they can be commoditized. Therefore, as a platform, Uber can take a 30% cut from each aggregated "commodity" (i.e., human users). Restaurants hate this, drivers hate it. They all try to bypass the platform by having users pay directly in cash.
Passengers (or at least I) understand that the platform's ability to implement a reputation system is its value, and we avoid cash transactions today.
You see a similar situation with Jupiter on Solana. Its early influence came from its ability to route orders across exchanges and provide users with the best price. Unlike restaurants and drivers, Jupiter aggregates venues for my WIF buy orders.
However, Substack and Spotify operate on different parallel tracks. Spotify pays up to 70% of its revenue to record labels. Last year, it paid $13.75 billion to rights holders out of $20.22 billion in revenue. For every dollar of revenue generated, only $0.04 ultimately reaches the artist. Substack, on the other hand, takes only a 10% cut from the fees I pay to newsletter authors. It neither commoditizes readers nor authors. It simply positions itself as a tool. Perhaps this is intentional—if it had pricing power, authors on the platform might leave directly.
Spotify, due to its partnerships with numerous record labels, is a loose supply-side aggregator to some extent. Substack positions itself as a demand-side aggregator without pricing power, relying on its reader base.
These applications each aggregate different sides of the market. But their ability to accumulate capital (or pricing power) depends on how tightly integrated they are within the ecosystem. We will soon see another version.
Web2-native aggregators leveraged two forces for massive user bases:
Moore's Law drastically reduced smartphone costs. India alone has 800 million smartphone users. Today, there are about 5.5 billion people online globally.
Internet costs decreased, and bandwidth costs fell rapidly.
In contrast, cryptocurrency, as an economy, has a much smaller TAM. By best estimates, about 560 million users have interacted with cryptocurrency. Last month, there were about 10 million active DeFi wallets. These are distinct economies with very different rule sets.
One is driven by attention, the other by the on-chain capital flowing within wallets. We often confuse the behaviors of the attention economy with those of the transactional economy. For example, user behavior on prediction markets is not the same as on Instagram.
Revisiting Aggregation
Three years ago, when I first wrote about aggregators, I believed blockchains reduced the cost of verification and trust. The initial promise of the internet was access—you could sit in New York in your pajamas and buy goods from Temu in Shenzhen. I thought then that blockchains would enable real-time verification and settlement with suppliers at minimal cost. Or so I predicted the industry would develop.
In 2022, I wrote:
"We believe blockchains will spawn a new class of marketplaces capable of instant verification of on-chain events. This will drastically reduce the cost of verifying intellectual property at scale, thereby creating new business models."
In the years since, new marketplaces have indeed launched. NFTs have traded about $100 billion in volume since then. Perpetual futures have liquidated about $14.6 trillion. Decentralized exchanges have cleared about $10.8 trillion. The understanding that the technology would be used to clear transactions with anonymous counterparties globally was correct.
But it must also be noted that OpenSea's volume has fallen from about $5 billion then to about $70 million this month.
Marketplaces emerge, evolve, and then die. Like many things in life. In the process, they leave us fuel for thought. Sid believes speculation in cryptocurrency is a feature, not a bug. All new markets are ambiguous in their early days; participants aren't sure what they are buying or why something has value. Novelty is built into the price. When rationality returns and valuations become efficient, the bubble becomes a memory of the past. Both NFTs and DeFi have gone through such manic cycles.
These bubbles were crucial for stress-testing and validating the monetary rails that ultimately underpin marketplaces like Hyperliquid.
This brings me to the experience in crypto over the past few weeks. In the Drift and Kelp incidents, about $578 million was hacked in the past three weeks. Over the past twelve months, about $1.7 billion has been stolen from DeFi and crypto protocols tracked by DefiLlama. Meanwhile, total protocol revenue across DeFi was about $3.42 billion during the same period.
In other words, for every $1 of DeFi revenue earned in the past year, about $0.50 was lost to hacks.
At the same time, we see more and more applications being released; the software itself has become a commodity. App Store submissions increased 84% this quarter compared to the same period last year. We see two forces at work simultaneously: more software competing for the attention of fewer users, while a few platforms dominate most of the revenue created in crypto.
Now we realize the industry loses $0.50 for every $1 it makes. The turn is sharp—but stay with me.
If we break down the $3.5 billion in revenue created through decentralized channels, you quickly see a pattern. About 40% comes from derivatives platforms, with Hyperliquid alone contributing about $902 million. The second largest category is decentralized exchanges, where the leader Uniswap generated about $927 million in fees. Lending platforms like MakerDAO rank third, with a combined revenue of about $500 million. What they have in common is that they are all capital-intensive businesses.
Unlike products that can be built with a few lines of code on a Saturday afternoon, they require the coordination of patient capital willing to accept the risks of these platforms.
This is where you realize the main difference between Web2 aggregators and Web3-native aggregators. Because blockchains are primarily tools for moving money and verifying that transactions follow the rule sets set by developers—they only become valuable when they can engage in capital-intensive operations. Perpetual futures exchanges can repeatedly put large sums of money to productive use within the same day. Lending platforms take a small cut from the large yields generated.
For example, Aave earned about $123 million from the $920 million in yield it generated over the past year. But such aggregators can only dominate the market if they can simultaneously have three key things:
- Supply side (liquidity)
- Demand side (users)
- Distribution
Hyperliquid is a unique beast in this regard. It has paid builders nearly $100 million in code fees, but the vast majority of its revenue comes from its own native frontend. It both keeps its best users and expands the surface area for new users to interact with the protocol.
But what's the logic behind this? One theory is that in Web3, distribution is a toll. Large protocols tend to own and retain their best users. You see this when you notice the revenue created by decentralized exchanges compared to on-chain aggregators that route orders.
On Ethereum, aggregators account for 36% of all DEX volume. On Solana, this figure can be as low as about 7% depending on the month. Kyber, 1inch, CoW, and ParaSwap have collectively generated only about $112 million in fees since launch. Meanwhile, Uniswap, which dominates most volume as a standalone exchange, has lifetime fees of about $5.5 billion. You see a similar phenomenon on Hyperliquid.
Builder code collectively accounts for only about 6% of Hyperliquid's cumulative $1.1 billion in revenue. MetaMask's deep integration on Ethereum earned $184 million in fees last year via swaps. Phantom generated about $180 million, but considering this is a massive ecosystem, it's just a small part. These products only work when built on top of a single protocol that has liquidity and economic activity.
These products can attract and retain users precisely because they have deep liquidity. From this perspective—capital in cryptocurrency is no longer a commoditized product. It is the most necessary ingredient. Vertical integration of capital gives participants more reasons to stay within the ecosystem. In such a system, capital begets more liquidity because it can be put to productive use.
Capital is not the moat; it is the result of vertical integration. Vertical integration is the moat; capital is just the byproduct.
It's important to note that this model only works when parked capital is not incentivized. Don't believe me? Look at any pre-launch protocol with an airdrop plan, and you'll understand, or look at the countless L2s struggling to generate value.
Any business engaged in capital aggregation is, to some extent, a target for hackers. Drift was targeted because it had about $570 million in TVL. KelpDAO was targeted because it held about $1.6 billion in restaked ETH. Hyperliquid's bridge holds about $2 billion in user deposits, making it one of the most valuable attack targets in the space. You can see similar situations with Ronin (~$625 million) and Nomad (~$190 million).
Because blockchain-native businesses require significant capital to generate value, we face a dynamic: to win, you must remain vulnerable until security mechanisms and mechanisms to freeze fund flows are in place.
Even if you have capital, massive TVL itself does not guarantee success. In an economy, idle or underutilized capital can become a liability when hit by a hack. This is why protocols try to differentiate themselves through the economic output they can generate, starting with niche use cases.
CHIP (the company behind USDAI) has issued about $100 million in loans this quarter, with a $1.5 billion pipeline moving forward; these higher-risk tranches will generate about 16% APY this year.
The highest-risk pool on Maple has an APY of about 15–20%, comparable to or higher than the current 12.6% APY on Aave's USDC pool. It aggregates borrowers who can utilize the protocol's liquidity to create economic output.
Naturally, Hyperliquid is the best example of a supply-side aggregator being able to put capital to meaningful use. Over the past year, Hyperliquid generated about $942 million in revenue with an average TVL of about $3.5 billion. In a very rough calculation, every $1 of capital parked on the protocol turns over about 285 times a year, generating about $0.30 in fees per $1 of TVL. Compare this to Aave, which generates about $0.05 in fees per $1 of TVL in its lending markets.
In a market where consumer preferences are not yet fixed and investor loyalty is low, capital flows to where it generates the best results. When considering hack risk, investors demand a risk premium. For now, perpetual futures exchanges are the only places that can repeatedly use idle capital on-chain and generate fees.
I initially thought Hyperliquid was just a supply-side aggregator, providing capital for users willing to trade on-chain. This has been my thesis. That's not wrong. But this thesis falls apart when you consider how it uses its token to incentivize vertical integration. But before we continue, let me explain how a vertically integrated ecosystem works.
Ticketmaster handles 70% of major live events in the US. It can take a 30% cut from your ticket to see Justin Bieber sing old songs at Coachella because it owns the venues, promotes the tour, ensures you can buy merch at the concert, and coordinates with sponsors. The 30% you pay is 15 times what Stripe charges to process online ticket fees. But you are willing to pay this premium because Ticketmaster is vertically integrated across the value chain.
You have a market illusion: artists, venues, and fans are all participants, but no one can challenge Ticketmaster's cut. The same goes for Apple's App Store. Apple curates the store, collects fees, ensures devices work, and brings millions of users already accustomed to the Apple payment "ding"—even if you subscribe to yet another app you'll never use.
Vertical Integration in Cryptocurrency
Protocols have begun to slowly implement the same logic.
In Web3, capital providers can be seen as commodities if there is no vertical integration making it easier for them to cooperate. Users will not have loyalty until the ecosystem is integrated into a cumulative experience that cannot be replicated elsewhere.
For Maple, this integration required years of experience dealing with hedge funds and market makers. For Centrifuge, integration includes securing nearly $1 billion from Grove for Janus Henderson's JAAA bond issuance. They are not capturing loose, abstract parts of the economy; they are providing a better product for end-users through vertical integration. In doing so, they create moats that are difficult to replicate overnight.
Maple's years of underwriting experience, or Centrifuge's moat as a trusted capital coordinator, are the moats in a world where capital and relationships are the only things difficult to copy.
Companies engaged in vertical aggregation may routinely hand off certain parts of the stack to third parties. Part of the reason is that there is no huge economic benefit in doing so. Maple owning custody or MetaMask issuing its own card might not bring huge profits compared to the capital they generate in swaps and credit underwriting.
However, when a business is growing exponentially, owning the entire stack is where competitive advantage is built. This is also part of the reason for M&A in the industry.
When a business is vertically integrated, you are not competing against a single product. The battle is over the combined experience the user gets. On Hyperliquid, once HIP-4 goes live, users can deposit for free (via Native Markets), participate in prediction market positions, and use that position as collateral to trade on its perpetual products. Its risk engine makes this experience possible. And it's worth mentioning that even in traditional finance today, this would be impossible without merchant banking.
Hyperliquid owns the users, the on-ramp, the risk engine, the trading interface, the liquidity, and the token issuance rights. For a new business to compete means fighting on six different fronts simultaneously.
For new applications launching, plugging into a small part of this is far better than building on a new protocol like Monad, which has only $2.6 billion in cumulative derivatives volume (spread across its five perpetual protocols).
Integrated ecosystems like Hyperliquid attract developers, more integrations, headlines, and happy token holders.
Exchanges are also seeing this shift. Coinbase acquired Deribit, has a custody business, co-issues USDC with Circle and earns from the reserves, has massive wallet infrastructure, and has on-ramps in over 100 countries. It also launched its own chain in pursuit of a vertically integrated experience. Admittedly, Coinbase might have been too early in pursuing retail users who clearly don't want to "mint" content on the blockchain or use Farcaster.
Coinbase's integration exists in a loose form but is hidden behind layers of bureaucracy, regulatory hurdles, and internal priorities. This is perhaps the main difference between open integrated systems and closed ones. As a ~$60 billion market cap exchange, Coinbase has little incentive to pursue fringe developers.
In contrast, Hyperliquid benefits by developing its core channels into the best trading venues while creating an ecosystem and creating value for the underlying token.
In this context, the token is part of the integration because it is the shared substrate that keeps these integrations alive and valuable. This is why the industry confuses tokenized protocols and tokenized businesses. The premise of a tokenized protocol is that third-party developers can easily build on it. It incentivizes people to direct value (down) to the token—often in the form of buybacks from the market.
Companies like Robinhood and Coinbase are powerful economic players, but they cannot replicate Hyperliquid's core owner-operator network.
Protocol airdrops ensure that those who own it are the individuals who contributed economically to it. They own enough tokens to drive value towards it. Hyperliquid commits to this cause by using 99% of its revenue to buy back tokens from the market. Imagine a public company using 100% of its revenue to buy back employee ESOPs. We might see an increased acceptance of capitalism.
This is why the industry is evolving—whether we like it culturally or not. Solana focuses on immutability, Ethereum on censorship resistance and open source, but you see the industry adjusting its ideology based on commercial realities.
Hyperliquid, while a beautiful garden, is a walled garden. To my knowledge, its source code is not open source. How its risk engine works is not verifiable. Maple's risk underwriting parameters are not public either. As a lender, I might not even know who underwrites the loans on USDAI.
Negotiating with Chaos
An economy cannot be built if $0.50 is lost to hacks for every $1 of revenue created. If founders are held responsible for hundreds of millions of dollars in TVL parked on their products, they will flock to AI. Whenever a hack occurs, we rush to hope stablecoins are frozen. And this in turn often means centralization.
Vertically integrated stacks ultimately need to sacrifice complete decentralization for economic progress.
This is not a new story on the internet. In the late 1990s, there was a strong dream of an open internet allowing free speech without consequences. Nazi memorabilia was auctioned on Yahoo until a French court intervened in 2000. Tim Wu delves deep into this theme in "Who Controls the Internet?". The story of the internet, or rather all human commercial networks, is that complete decentralization gives way to a milder version, sacrificing some control for economic interaction.
We accept diluted versions of the original vision so that commerce can scale, because without dilution, chaos ensues.
This massive energy expansion is reflected in how we describe those eras: the "wild" west, the internet "bubble." Perhaps cryptocurrency is undergoing a similar expansion and convergence of energy. I explored these themes in depth last year in the following articles.
What Does This Mean for Founders?
Look at the data from MetaMask and Phantom. These businesses make more money than most L2s because they are downstream of ecosystems with huge economic output. Building bridges and exchanges where there is no liquidity and no users is no longer a viable business model. Especially when it comes with the pain of hacks. You should build where there is liquidity and users today.
Imitating a vertically integrated product overnight might be impossible, but you can build on top of it.
Operating systems went through a similar pattern. When BlackBerry declined and iOS became dominant, developers had to choose where to build. We see a similar situation in crypto. Only this time, capital incentives might keep developers blind for longer.
Platforms and protocols on the internet are very similar to this. We might not like the rules, but they keep things functioning.
In the era of vertical integrators, we may increasingly agree to some rules so that our money stays with us, and the economies we invest significant time in can continue to scale. The trend points in this direction. Stablecoins, RWA, perpetual futures exchanges with closed risk engines, lending platforms with unknown risk underwriters, and off-chain RFQ products like Derive all point to the same trend.
That is, vertically integrated capital aggregators willing to sacrifice the dream of complete decentralization for progress.
