On November 11, 2026, at 09:27 (UTC+8), ChainCatcher published an essay by Spencer Applebaum, general partner and co-head of venture investing at Multicoin Capital, compiled by Gu Yu. It is an investor’s framework for where value might stick in DeFi, not a customer disclosure, a reserve report, or an audit. Multicoin writes that it fully intends to invest where recent, unnamed SEC guidance and rule proposals have mapped a compliant path for tokens to capture value. That stake belongs in front of the argument.
The question worth keeping is not whether a model can emit an exchange. It is who controls the assets, who is owed what, and who bears the loss when a position fails. Applebaum’s answer is that those positions — distribution of balances and order flow, a balance sheet that can absorb losses, and data others cannot see — are not what cheaper software produces.
What he says got cheaper
Applebaum writes that AI is making competition easier across almost every part of the software industry, but that financial services will be one of the hardest businesses to commoditize. His comparison, stated as industry context, is that five years before the essay, building a DEX or a lending protocol meant raising $3 million, hiring five engineers, and spending six months on a minimum viable product. By November 2026, he says, a founder can assemble most of that in a weekend with Claude Code. Ask a frontier model to build a perpetual-futures exchange and, out of the box, it returns a matching engine, a risk engine, a trading interface, an API, and most of what else is needed.
The valuable remainder, in his account, has little to do with the software. A loyal customer base, a large balance sheet, and proprietary data plus underwriting ability each take years of relationships, a track record, and trust. A model can write a lending protocol in an afternoon. It cannot conjure $10 billion of stablecoin deposits into a contract, or sign 28 million customers for a neobrokerage. That 28 million figure is the same count he later attaches to Robinhood’s funded accounts.
He calls the combination strongly positive for DeFi. Over the next decade, he expects trillions of dollars of financial activity to move onchain, with AI cutting the cost of products built around that activity, while the inputs that make a financial business defensive do not get easier to copy. He also states a standing belief that DeFi will keep taking share from centralized crypto venues and, in time, from traditional finance. He points to DEX share versus centralized exchanges at historical highs and still rising, citing The Block and DefiLlama. The ChainCatcher text does not print the percentages behind that chart. He separately points to real use of tokenized assets — Treasuries, private credit, and equities — and says growth is not only crypto-native: existing institutions, fintech firms, institutional asset managers, and traditional brokers are building onchain and bringing clients with them, citing rwa.xyz. That chart’s series is not in the article text either.
His caution matters more here than the adoption claim. Today’s blue-chip DeFi protocols, he writes, will not necessarily be the ones that capture the growth. As software and data analysis get cheaper, value moves toward whoever controls scarce inputs the software cannot create.
A framework the firm says it has already revised
Applebaum presents the piece as the next pass at how Multicoin thinks about value capture. Earlier, the firm looked at application tokens capturing value through work tokens and burn-and-mint equilibrium models. Later it focused on defensibility and on state that cannot be forked: anyone can fork open-source code, but not liquidity or network effects. More recently it argued that protocols capture the most value when some risk has to be managed. He says each of those arguments has, in some measure, been borne out.
This iteration names three moats: distribution, capital, and proprietary data. A fourth factor, willingness to bear risk, works differently. He treats it as a wedge, not a moat, and returns to it at the end.
Distribution: control of where balances and orders go
He defines distribution narrowly. It is control over where order flow or balances are sent. Not all of it is equally sticky. Funded accounts that are hard to move are a real barrier. An interface alone is easier for a competitor to hit.
He sets this against Joel Monegro’s 2016 essay “Fat Protocols,” which argued that blockchains would invert the internet’s pattern — value at Google, Facebook, Amazon, and other applications — and send most value to the base protocol. Applebaum says the evidence points the other way, and that the trend is accelerating. As protocols get cheaper to launch, value accrues to whoever holds the customer relationship and can route users and balances to the highest bidder. Sometimes that party is an application. Sometimes, as with Hyperliquid, it is a chain that is also an application.
The issuer’s reserve yield versus the venue that holds the balances
On stablecoins, the essay separates the issuer’s reserve business from the distributor’s leverage.
Issuance, he writes, has historically been an excellent business if the issuer reaches scale and rates are not near zero. Tether and Circle hold dollar reserves, earn interest on them, and in most cases pay holders nothing. Tether alone, he says, has since 2021 earned more than $20 billion of interest income at margins near 100 percent, citing DefiLlama. That is issuer income as he reports it. It is not a statement of what token holders are owed. Reporting that an issuer holds dollar reserves and earns interest does not establish a liability schedule, an audit, solvency, or any holder claim on the interest.
Hyperliquid is his example of a venue that can demand a cut because the balances are there for its market. The essay says the platform had more than $6 billion of USDC posted as derivatives collateral, and that those balances generate considerable reserve income for Circle and its distribution partners. At some point, he writes, Hyperliquid recognized that it was the reason the balances existed, and therefore had leverage: if Circle would not share the economics, Hyperliquid could replace USDC with a stablecoin whose issuer would, or in principle issue its own.
Earlier in 2026, relative to the November publication, Hyperliquid used what the essay calls its Aligned Quote Asset (AQA) framework to reach a deal with Coinbase. Under that deal, about 90 percent of the USDC reserve yield on Hyperliquid flows to the protocol’s Assistance Fund, which uses it to buy HYPE, rather than to Circle or Coinbase. The essay says the vault held $5.5 billion of USDC — the base tied to that 90 percent flowing toward HYPE — and that this was, at the time of writing, about a $185 million revenue line for Hyperliquid. Both USDC figures are his: more than $6 billion as derivatives collateral on the platform, and $5.5 billion in the vault associated with the yield split. The text does not reconcile them, and this reading will not.
The obligation sits in front of the revenue line. Hyperliquid, in his account, does not maintain the plumbing: not mint and redeem, not vault management, and not the regulatory licenses. It has distribution, and he argues that was enough to demand more of the economics. USDC visible onchain as collateral is not Hyperliquid owning the dollars, the reserve portfolio, or a legal claim on yield. The split he describes is a negotiated arrangement with Coinbase. The Assistance Fund’s use of that yield is to buy HYPE, not to pay USDC holders.
Front ends price the route. They do not run the risk engine.
Wallets and front ends are the other side of the same split. Non-custodial wallets, he writes, struggled for a long time to monetize, then recognized that trading was crypto’s most profitable business and, a few years before the essay, made it a first-class product. On trades, MetaMask charges 0.875 percent. Uniswap’s average trading fee over the prior twelve months was 0.13 percent, and he says most of that value goes to liquidity providers rather than to the Uniswap protocol. The wallet, in short, earns far more than the venue that executes.
The exchange still has to attract liquidity providers, maintain complex infrastructure, and compete on execution. MetaMask has to win and keep the customer. If a DEX does not offer a good price, MetaMask can route elsewhere. With enough order flow over time, he argues, a front end can internalize a large part of that flow and keep more of the economics.
He sees the same pattern in perpetuals. Phantom lets users trade perps on Hyperliquid and charges an additional 5 basis points on top of Hyperliquid’s 4.5. Hyperliquid runs the risk engine, the matching, and the liquidity. Phantom, he writes, earns more only because it owns the interface and the user.
He marks a limit. Among Hyperliquid builders, the essay describes a new red-ocean market already in a price war as of November 2026, with teams cutting fees to improve retention. The front-end layer is not exempt from commoditization. More generally, he calls crypto wallets somewhat fragile because switching costs are very low: a user can connect a private key or seed phrase to another wallet and, within minutes, reach the same funds through a new interface. Leaving a bank or a fintech firm is harder. The user has to open a new account, complete KYC, move money over traditional rails, reconnect payroll, and may have to update other connected services. A balance that moves with a seed phrase is not the same relationship as a funded account tied to those rails.
A funded account can threaten to build the venue
Robinhood, in the essay, has 28 million funded accounts. As of the piece, Uniswap was the main DEX on Robinhood Chain, and Robinhood did not object. Applebaum is explicit that he is not saying Robinhood will certainly launch its own DEX. Owning liquidity has not historically been Robinhood’s primary goal, though he notes that Robinhood had started routing substantial flow to its own prediction market, Rothera, rather than to Kalshi. The narrower point is that nothing structural stops Robinhood, as build costs fall, from shipping a DEX protocol and sending users there. With its reach and then-current coding tools, he writes, it could stand up one of the larger layer-2 DEXes in DeFi almost overnight. Building the protocol is no longer the hard part. Attracting users, capital, and proprietary data is.
He describes a waterfall taking shape. Front ends such as Phantom are taking more of the economics from venues such as Hyperliquid, and venues are taking more from stablecoin issuers such as Circle. In each case the business with distribution has pricing power over the infrastructure below. How much a layer keeps depends on how easily users can leave. Hyperliquid, in his account, secured about 90 percent of the reserve yield. Front ends on Hyperliquid had already competed fees down to low single-digit basis points.
Capital: the input a model cannot mint
A natural conclusion of that waterfall, he says, is that headless protocols and back-end infrastructure are at a disadvantage. In some cases they are. There are other moats besides distribution, and he calls capital one of the most important.
Capital is the lifeblood of these markets. Traders need it for leverage, market makers need it to keep quoting, and lending protocols need deposits in order to lend. AI can write code and automate workflows. It cannot create the money those systems run on. When something breaks — he names Aave’s KelpDAO episode and MakerDAO’s Black Thursday, without retelling either — someone has to take the loss. A protocol interface is not that someone.
Liquidity, in his telling, is only partly defensive. Traders go where liquidity is, and liquidity follows them; he calls this a log(n) network effect. Hyperliquid’s open interest is a real advantage, but that capital belongs to users who can move it. DeFi has already seen liquidity shift from one protocol to another, often on token incentives and vampire attacks, and often leave when the incentives dry up.
He names two forms of capital that can support longer-lived value.
One financing pool, or collateral posted at every venue
A prime broker in traditional markets provides leverage, finances trades, and lets a hedge fund use a single pool across markets and strategies instead of posting collateral at each venue. Applebaum calls the business enormous. Industry prime-brokerage revenue, he writes, rose from about $15 billion in 2020 to about $27 billion in 2024, behind growth in hedge funds and in multi-manager pod shops that depend heavily on financing.
He thinks the DeFi opportunity may be larger. Traditional prime brokers mainly serve large institutions, because smaller clients are operationally expensive. Smart contracts, he argues, could automate much of that work and offer it to anyone from a person trading $500 to a pod inside Millennium managing hundreds of millions of dollars. That is a product thesis. The essay does not present a broker already operating across that span.
The inefficiency he describes is specific to the time of writing. Traders typically had to post collateral separately on Hyperliquid, Aave, Lighter, Morpho, Variational, and Kamino. He calls that extremely capital-inefficient, and expects it to worsen as tokenized equities, Treasuries, private credit, and other real-world assets move onchain. A fund long a tokenized stock in one venue and short an equity-index perpetual in another should not, where the risks partly offset, have to collateralize both in full.
The software to track that netting is getting easier. The scarce pieces are the balance sheet that finances the positions, a lower cost of capital, and the risk-management skill to avoid liquidation. Over time, he argues, a prime broker with cheaper capital can offer better financing terms, attract more clients, and use that scale to finance itself more cheaply. The business, in his view, ought to show strong returns to scale. Better financing terms are a spread on a balance sheet. Nothing in the essay turns them into a customer yield or a proof of reserves.
Whose dollars stay when the trade fails
The second use of capital is agreeing to backstop risk. This continues an earlier Multicoin argument about DeFi DAOs that manage risk: tokens tied to protocols that do not manage risk are easier to fork out than tokens tied to protocols that do.
Hyperliquid’s HLP vault, as described here, provides liquidity by making markets and by taking over liquidated positions. Depositors share the profits and also absorb the losses. In March 2025, HLP lost about $4 million while closing a large ETH position. A later JELLYJELLY episode, the essay says, exposed the vault to further risk; it states no dollar figure for that episode. Applebaum treats both as evidence of why a derivatives DEX’s insurance fund matters. Aave’s Umbrella is the parallel he offers: users can earn yield in exchange for backstopping bad debt. That yield is compensation for bearing someone else’s default. It is not a reserve attestation, and it does not establish that the vault, the protocol, or anyone else has been audited.
Anyone can fork the HLP or Umbrella contracts. They cannot fork the capital sitting in them. His caveat is that this capital is third-party money. If depositors find a better yield, they may leave. The counterargument he offers is overlap between token holders and depositors. A HYPE holder might keep some dollars in HLP so the exchange is safer and the token investment is protected. That is a hypothesis about sticky capital, not a measured retention rate.
The more durable version, he writes, is the protocol’s own balance sheet, built from retained earnings. That capital does not leave in pursuit of a newer, shinier yield or a points program.
Binance is his illustration, and he marks the lesson as early. The essay says that in a hack on October 29 — recent, relative to a piece dated November 11, 2026 — the exchange lost $6.3 billion, made up the loss with SAFU reserves and company funds, mainly the company funds, reopened withdrawals within days, and saw only a net 3 percent of assets leave. The reading he offers, while saying it is still early, is that the perceived balance sheet helped stop a bank run. A perceived balance sheet and a reported net outflow are not an audit. They do not establish solvency, the legal priority of customers against SAFU reserves versus company funds, or the protections a given depositor actually has. The essay’s own split still matters: company funds, not a wallet balance, did most of the filling. On-chain balances alone do not show who owed the hole or who paid it.
As DeFi moves further into private credit, tokenized equities, and other real-world assets, he expects demand for insurance and risk capital to keep growing. Protocols that actually supply and manage capital will, he concludes, capture far more value than protocols that only deliver software.
Proprietary data is an underwriting file, not a block explorer
The third moat is information, with a limit he states first. DeFi was built on the idea that financial data should be public, and that works well for crypto-collateralized lending and for spot trading. As more real-world assets move onchain, he expects more businesses built around information that is not public. The moat is that data combined with the ability to underwrite from it.
Curators on Morpho, Kamino, and Euler are where he says this appears first, and where it has mostly failed to produce fees. They assess risk on behalf of depositor funds and, at the time of the essay, oversaw $9.5 billion. Inside vault smart contracts they are granted limited control: which lending markets receive funds, what the risk parameters are, and monitoring of collateral. Limited control is not ownership of the deposits, and the visible assets are not a statement of legal title.
Despite $9.5 billion, he says they generate very little revenue, especially against traditional asset managers with wider discretion and wider responsibility. Many curators charge nothing. Steakhouse Financial, in his account the largest DeFi risk curator by assets, takes 22 basis points. The average curator take rate is 19 basis points.
He offers two partial causes and a larger third. Risk-adjusted yields in DeFi were, as of the essay, unattractive, very likely in part because of idle capital sitting in Aave — the parenthetical cause is his. Morpho and Kamino had homogenized the vault layer and sit between curators and depositors, which returns him to distribution: the platform facing the depositor can compress the curator’s fee. The larger problem is that curators mostly work from the same data. Borrow against ETH or BTC and the price, the liquidation rules, and the collateral are public. AI makes the risk model easy to copy, so curators compete on judgment — for example, on taking risk, a large part of which is already predefined by algorithm — and on track record. Over a long enough span, he says, that will inevitably put pressure on fees.
He contrasts OnRe, described as an onchain reinsurer. It takes capital from investors and uses it to underwrite real-world insurance risk. That requires historical claims data, loss probabilities, and the insurance contracts themselves, plus people who understand the file. None of this is on a block explorer. A frontier model with no access to the reinsurance book cannot underwrite it, however intelligent the model is. OnRe, he writes, can charge a 75 basis point management fee plus 20 percent of yield. On then-current returns, that was about 3 percent a year, 16 times the average DeFi risk curator.
These are different businesses with different risks, and he notes that part of the fee premium reflects illiquidity and the possibility of loss. That is the point he wants kept. A firm with proprietary data, underwriting skill, and relationships can charge much more. A firm with the same data and the same tools as everyone else will, very likely, end up a commodity. The 16-times figure is his comparison across those businesses. It is not a claim that curator fees should rise to reinsurance fees, and it does not make OnRe’s book visible, audited, or comparable in losses to a Morpho vault.
Chains, applications, and what scarcity is left
A cynical reading he entertains is that infrastructure tokens, L1s and L2s included, are in a difficult spot. Blockchains used to hand a new DeFi application two of the three moats: distribution and capital.
Historically, he writes, a chain had a distribution advantage because users kept assets on the chain where they were most active. Building on Ethereum reached Ethereum users. Building on Solana reached Solana users. That mattered for anyone starting Uniswap, Compound, Aave, Kamino, or MakerDAO with no existing customers, no assets, and no financial relationships. Launching where users already were made the start much easier.
Interoperability, in his account, is weakening that. A user onboarded through the fomo app can trade any asset on any chain without having to think about which chain it is. He expects more new users to enter DeFi through fintech firms such as Revolut, Nubank, and Robinhood, which already have tens of millions of customers, and which do not require an already funded wallet in order to be reached. “Tens of millions” is his group characterization. The only precise funded-account count in the essay is Robinhood’s 28 million.
Chains also supplied capital through ecosystem wealth effects. As SOL and ETH rose, users had more to deploy into applications on those chains. Capital then tended to stay, because bridging was painful and users were tribal. He says bridging is easier now, and that capital is arriving through stablecoins, tokenized assets, and institutions, so a new protocol depends less on any single token’s wealth effect. Real-world-asset issuers also want to launch on every chain. He points to Securitize, Superstate, and Ondo as cases already visible.
On data, he says there was never much of a moat. Almost all onchain financial information is public, so neither chains nor applications ever had much proprietary data to defend. AI is also eroding another historical blockchain advantage, the developer ecosystem. One of Ethereum’s largest edges over newer chains such as Solana or Aptos was mature tooling, audited libraries, and integrations. As models make it easier to port applications across virtual machines, that edge matters less.
General-purpose chains, in this telling, are losing their advantages in distribution and capital, and they did not start with much of a data moat. They compete mainly on execution and settlement. Blockspace is abundant, and fees keep falling. He does not conclude that a chain cannot be valuable. He cites a summit talk by Tushar a few years before the essay, which described L1 tokens as capital assets, commodity assets, and stores of value. The number that follows is a conditional illustration, not a statement of Solana’s results: if Solana eventually processed a billion payments a day at one cent each, that would be $3.65 billion of fees a year. He says transaction volume continues to grow, pointing to a Blockworks Research chart. The article text does not include that series. A hypothetical fee run-rate is not revenue. A rising transaction count is not a reserve, a liability schedule, or evidence that a token captures the fee. His conclusion is narrower: chains may be valuable for reasons quite different from the ones investors first expected.
For applications, the firms he expects to capture the most value are those with at least one of the three moats. Distribution, through a direct customer relationship that is hard to migrate, with Robinhood or Hyperliquid as his examples. Capital, through a balance sheet that can absorb losses. Information, through the ability to underwrite risks others cannot. All three, he says, are much harder to copy than software, however good the models become.
He does not think it matters whether the value sits in equity or in a token. His view is that it flows to both. What matters, on his standard, is whether a business — token or share — controls something scarce, has pricing power because of that, and can turn it into an investor return regardless of the instrument. That is an investment filter. It is not a finding that any named token has such a path, and nothing in the essay is a recommendation to buy or sell one.
As more financial activity moves onchain, he expects the gap to widen between protocols that hold a moat and protocols that are only software. Multicoin, which he describes as one of the largest investors in the sector, spends a lot of time deciding which projects in the stack have these moats and can keep them.
Incumbents already hold the expensive inputs
A reasonable conclusion, he grants, is that incumbents win. Robinhood has the customers. Goldman Sachs has capital that may exceed anything a future DeFi prime broker is likely to hold — his comparison, not a balance-sheet extract. Moody’s has the data and the licenses. If software is becoming free and the moat is the expensive part, firms that already own the moat should be favored.
The startup advantage he keeps is a willingness to take risks incumbents will not. Part of why that opening exists, he writes, is that AI cuts both ways. The same tools that would let Robinhood ship a DEX in a weekend let a three-person startup do it. As the cost of building a credible product falls, the real opportunity is markets and customers the incumbent will not chase.
His examples are his selection, not a census.
Ethena built a synthetic dollar around delta-neutral hedging. The basis trade, he writes, is a strategy traditional firms clearly understand but could not or would not bring to crypto exchanges. By the time of the essay it had scaled to $5.4 billion of assets under management. Assets under management are not reserves matched to a liability, and a strategy label does not say who is owed the dollars if the hedge fails. The essay does not set out that chain of obligation.
Pump.fun and fomo, he writes, emerged with new trading experiences in a new asset class, memecoins, and built substantial distribution around that wedge. He calls them two of the largest success stories of this crypto cycle. He does not define success as profit, solvency, or retained customers.
Hyperliquid entered a derivatives market dominated by Binance, Bybit, and OKX and, he says, took share quickly. It has also listed perpetuals on pre-listing companies such as OpenAI and Anthropic, which traditional exchanges will not do. Those markets, he writes, drew new traders and liquidity, visible in wallets created on the day of the SpaceX IPO. He cites HyperTracker and gives no wallet count in the text.
On capital, Project 0 finances DeFi collateral that traditional prime brokers will not accept as lending collateral. On data, USD.ai can underwrite loans backed by GPUs, an asset class in which he says banks have almost no experience.
The common thread he draws is that these businesses went after markets incumbents were not serving, whether because of regulation, operational complexity, reputational risk, or because the opportunity looked too small. For startups, he writes, there will always be openings of that kind for asymmetric competition, as seen across crypto’s history and as he expects to keep seeing.
That is why he calls AI and DeFi together compelling. AI is significantly lowering the cost of building financial products. DeFi lets a startup reach customers and obtain capital without rebuilding the entire financial stack. Risk appetite itself, he is careful to add, is not a durable moat. Regulation catches up, and once a market is proven, incumbents follow. His example is spot bitcoin markets. A startup does not need a moat on day one. It needs an opening the incumbent will not take, and enough time to turn that opening into one or more of the three moats, and then to compound from there. “Compound” is his word for business durability, not a stated return.
What the close claims, and what it does not settle
The essay’s summary restates the bet. Over the next decade he expects trillions of dollars of financial activity onchain, and AI will make products serving that activity easier to build than ever. Software can be homogenized without homogenizing what makes a financial firm valuable. Customers, capital, and proprietary data remain hard to obtain, and the businesses that control them will have pricing power. That is why Multicoin is bullish on the next generation of DeFi companies. As AI developer tools spread, startups can contest incumbents faster and more cheaply than before, and a new firm that establishes one of these moats can, in his view, build an extremely durable business. Code can be forked and generated with the latest and strongest model. None of the moats can simply be generated. That, he says, is the opportunity as a crypto investor.
Read as a custody ledger, several things stay explicitly unsettled.
A wallet balance, an open-interest figure, or a curator’s assets under management shows where units are visible. It does not show legal ownership, a customer’s claim, or a promise to make a loss whole. Tether’s reported interest income is income to an issuer that, in most cases, does not pay holders. It does not establish the liability side, an audit, or solvency. Hyperliquid’s roughly 90 percent share of USDC reserve yield is a commercial split attached to distribution; minting, vault management, and licenses sit elsewhere on his account, and the yield he describes is used to buy HYPE. HLP depositors are paid to bear losses and, in March 2025, bore about $4 million of them. Umbrella is the same bargain. Binance’s use of SAFU reserves and, mainly, company funds is the essay’s evidence about who filled a stated hole, not evidence that on-chain balances were the cushion. Curator fees of 19 or 22 basis points are prices for limited parameter control over public collateral, not proof of underwriting skill. OnRe’s higher fee is tied, in the essay, to an off-chain book outsiders cannot see on a block explorer. None of these figures is a yield to underwrite from, and none is a view on what any token or share is worth.
Original article, published November 11, 2026, 09:27 (UTC+8): ChainCatcher.