Crypto Versenet

A Weekend DEX Is Not a Moat: Multicoin’s November 2026 Argument

On November 11, 2026, PANews published a Chinese compilation—credited to 大钳子—of an essay by Spencer Applebaum, an investment partner at Multicoin Capital. What follows tracks that text. It is an investor’s framework for how value might be captured in DeFi, not a set of measured market prints, and it is not a case for owning any token.

The story being told

Applebaum’s claim is that frontier AI has collapsed the cost of shipping financial software, including a decentralized exchange, while the scarce inputs in finance have not become easier to copy. Five years before the essay, he writes, a DEX or lending protocol meant raising $3 million, hiring five engineers, and spending six months on a minimum viable product. By November 2026, he argues, a founder using Claude Code could do most of that work in a weekend. Ask a frontier model for a perpetual-futures exchange, he says, and it can return a matching engine, a risk engine, a trading interface, an API, and most of the basic functions.

He treats financial services as unusually hard to turn into a commodity, because the valuable part is not the software. A loyal client base, a large balance sheet, and proprietary data plus risk assessment take years of relationships, a track record, and trust. A model might draft a lending protocol in an afternoon. It cannot, in his illustration, place $10 billion of stablecoin deposits into a contract, or recruit 28 million customers for a new digital bank.

Multicoin’s belief, offered without a forecasting method in the essay: over the following decade, trillions of dollars of financial activity move on-chain; AI makes the products that serve that activity cheaper to build; the traits that set a financial firm’s value do not get cheaper to replicate. The firm also says DeFi will keep taking share from centralized finance and eventually span traditional finance. That is the narrative. The material under it is a series of examples, plus charts the prose does not quantify.

Applebaum says this extends earlier Multicoin lines of thought: application tokens capturing value through work tokens and burn-and-mint equilibrium; then “unforkable state,” the idea that open-source code can be copied but liquidity and network effects cannot; then the claim that protocols capture more when they take and manage a risk. He says those views were partly borne out.

What he points to as adoption

He writes that DEX share versus centralized exchanges had reached a historical high and was still growing, citing charts from The Block and DefiLlama. The article text does not state the percentage. He also says tokenized assets—U.S. Treasuries, private credit, and equities—were finding real use, citing rwa.xyz, and that the push was no longer only from crypto-native users. Traditional financial institutions, fintech companies, institutional asset managers, and traditional brokerages were building on-chain products and bringing their own clients. No dollar total from that chart appears in the prose.

His caution cuts against reading “DeFi” as one beneficiary. Today’s leading protocols, he says, are not automatically the ones that capture the growth. As the cost of software and of data analysis falls, he expects value to move toward whoever holds resources software cannot create.

He names three moats—distribution, capital, and proprietary data—and a fourth factor, risk tolerance, which he says works differently: a way into a market, discussed at the end, not a finished barrier.

Distribution: control of flow, not of code

He starts from Joel Monegro’s 2016 “fat protocols” thesis. On the internet, most value accrued to applications such as Google, Facebook, and Amazon; blockchains were supposed to invert that, so base protocols would capture most of the value. Applebaum says events have run the other way, and faster. As protocols get cheaper to build, value shifts toward whoever holds the customer relationship and can send users and funds to the venue that pays them most. Sometimes that party is an application. Sometimes, as with Hyperliquid, it is a blockchain that is also an application.

His definition is narrow: control over where order flow or balances go. He adds that not all distribution is equally sticky. Accounts that already hold funds and are hard to migrate are a real barrier. A user interface alone is easier to contest.

Stablecoin yield

At scale, and when interest rates are not near zero, stablecoin issuance has historically been a very profitable business, he writes. Tether and Circle hold dollar reserves and earn the interest while, in most cases, not paying holders. Tether alone, he says, has taken in more than $20 billion of interest income since 2021, at margins near 100 percent, citing DefiLlama.

On Hyperliquid, he writes, more than $6 billion of USDC sits as collateral for derivatives. Those balances generate reserve yield for Circle and its distribution partners. His argument is that the venue created the balances and therefore has leverage: if Circle will not share, Hyperliquid can replace USDC with a stablecoin that will, or in principle issue its own.

Earlier in 2026, he writes, Hyperliquid reached an arrangement with Coinbase under an Aligned Quote Asset (AQA) framework. About 90 percent of the USDC reserve yield on the platform would flow to the protocol’s Assistance Fund and be used to buy HYPE, rather than remaining with Circle or Coinbase. On $5.5 billion of USDC in the treasury, he says that 90 percent share was then worth about $185 million of revenue to Hyperliquid. The venue does not, in his account, run minting and redemption, treasury management, or the regulatory licenses. It has the distribution channel, and that was enough to demand more of the economics. These figures are his description of the arrangement; they are not recalculated here.

Wallets and frontends

Non-custodial wallets, he recounts, long struggled to make money, then noticed that trading was crypto’s most profitable business and made it the product. MetaMask’s fee is given as 0.875 percent. Uniswap’s average fee over the prior 12 months is given as 0.13 percent, with most of Uniswap’s fee going to liquidity providers rather than the protocol. On that comparison, the wallet earns far more from the trade than the venue that executes it. The exchange must attract liquidity providers, maintain complex infrastructure, and compete on execution quality. MetaMask has to win and keep the relationship. If a DEX is not competitive on price, the wallet can route elsewhere, and a larger order flow can be internalized over time.

Perpetuals, he says, were starting to look similar. Phantom lets users trade perpetuals through Hyperliquid and charges an extra 5 basis points on top of Hyperliquid’s 4.5. Hyperliquid runs the risk engine, the matching system, and the liquidity. Phantom, holding the interface and the customer, takes more of the revenue in his telling.

He does not treat that layer as safe. Builder teams in the Hyperliquid ecosystem were, by the time of the essay, in a crowded price war, cutting fees to retain users. Even a frontend, he concludes, can be commoditized. Crypto wallets are also easy to leave: import a private key or seed phrase and the same funds are available through another interface within minutes. Leaving a bank or a fintech service is slower—a new account, identity checks, a transfer over traditional rails, payroll re-linking, and updates to other attached services.

Fintechs, including a path he does not predict

Robinhood, he notes, had 28 million funded accounts—the same scale as the “new digital bank” illustration in the opening. Uniswap was the main DEX on Robinhood Chain, and Robinhood was not objecting. His point is capability, not a forecast. Nothing in the cost of software, he writes, stops Robinhood from building its own DEX and sending users there; with that reach and current programming tools, it could stand up one of the largest layer-2 DEXes in DeFi almost overnight. He explicitly says he is not claiming Robinhood will do this. Historically, owning liquidity has not been the company’s first objective. What he does report is that Robinhood had begun shifting substantial trading flow from Kalshi to its own prediction market, Rothera.

The line he wants kept: building the protocol is no longer the hard part. Attracting users, capital, and proprietary data is.

The allocation chain he sketches runs upward in pricing power. A frontend such as Phantom takes more of the economics from a venue such as Hyperliquid; the venue takes more from an issuer such as Circle. Whoever holds distribution has more power to price the infrastructure underneath. How much each layer keeps depends on how easily users leave. Hyperliquid’s capture of about 90 percent of the reserve yield sits next to frontends in its own ecosystem competing fees down toward single-digit basis points.

Capital: code does not post margin

Protocols and back-end infrastructure without a direct user door are, he says, at a disadvantage—but not in every case, because capital is a separate moat. Traders need it for leverage, market makers need it to keep quoting, and lending protocols need deposits to make loans. AI can write code and automate workflows. It cannot create the money those systems run on. When something breaks—he names Aave’s KelpDAO incident and MakerDAO’s Black Thursday, without retelling them—someone has to bear the loss.

He separates ordinary liquidity from capital that can defend a business. Liquidity, as he describes it, has a logarithmic network effect: traders go where it is deep, and liquidity follows traders. Hyperliquid’s open interest is a real advantage, but the capital belongs to users, who can move it. DeFi has already seen liquidity migrate from one protocol to another, often on token incentives and “vampire” attacks, and leave again when the incentives stop.

Two forms, he argues, can create longer-lived defense.

A prime broker that can net offsetting risk

Prime brokers in traditional markets provide leverage and financing, and they let a hedge fund use one pool of capital across markets and strategies instead of posting collateral at every venue. He calls the business very large. Driven by hedge-fund growth and by multi-manager “pod shop” models that depend on financing, industry prime-brokerage revenue rose from about $15 billion in 2020 to about $27 billion in 2024. He thinks the DeFi opportunity may be larger, because a traditional prime broker mainly serves big institutions: small clients cost too much to operate. Smart contracts, he argues, can automate much of that work and extend it from a person trading $500 to a team managing hundreds of millions of dollars at Millennium.

Today, he writes, a trader often has to post collateral separately on Hyperliquid, Aave, Lighter, Morpho, Variational, and Kamino. That is extremely inefficient in its use of capital, and he expects it to get worse as tokenized equities, U.S. Treasuries, private credit, and other real-world assets move on-chain. A fund that is long tokenized stock on one venue and short an index perpetual on another should not, where the risks offset, be required to post full collateral on both.

The software to administer that business is getting easier. The hard part is a balance sheet large enough to finance the positions, access to cheap funding, and risk management that avoids a large loss. Over time, a prime broker with a lower cost of funds can offer better terms, attract more clients, and use scale to cheapen funding further. He expects significant economies of scale.

Who actually takes the loss

The second use of capital is backstopping risk, tied to earlier Multicoin writing on DeFi DAO risk management. A protocol that manages risk, that earlier view held, has a token that is harder to replace with a fork than one that does not. Hyperliquid’s HLP vault supplies liquidity by making markets and taking on liquidated positions. Depositors share profits and must also take losses. In March 2025, HLP lost about $4 million closing a large ETH position; a subsequent JELLYJELLY incident, he writes, exposed it to further risk. Both events, in his view, showed why a derivatives DEX needs an insurance fund. Aave’s Umbrella is the parallel he offers: users backstop bad debt in exchange for yield.

Anyone can copy the HLP or Umbrella contracts. They cannot copy the capital inside them. The weakness he names is that this capital comes from third parties. If depositors find a higher yield elsewhere, they can leave. A partial reply, as he frames it: some capital may stick because token holders and depositors overlap. HYPE holders might leave some dollars in HLP to make the exchange safer and to protect their token. The form he calls most durable is a balance sheet built from retained earnings, which does not leave for a newer yield or a points program.

Binance is his illustration of why that matters. After a hack on October 29—presented in the essay as recent, and given there without a year—in which the firm lost $6.3 billion, Binance covered the loss with its SAFU reserve and with company funds, mainly the latter, and reopened withdrawals within days. Net assets that left were 3 percent of total assets, he writes. He says it is too early to conclude, but that trust in the strength of the balance sheet may have helped avoid a run.

As DeFi moves into private credit, tokenized equities, and other real-world assets, he expects demand for insurance and risk capital to keep growing. Protocols that actually provide and manage capital, he concludes, will capture more value than protocols that only provide software.

Proprietary data, and why public ledgers are a weak barrier

The third moat is information, with a premise he flags. DeFi was built on the idea that financial data should be public, which works for crypto-collateralized lending and for spot trading. As more real-world assets come on-chain, he expects more businesses built around information that is not public. The moat is proprietary data combined with the ability to underwrite risk using it.

Curators on Morpho, Kamino, and Euler assess risk and manage depositor funds, a combined $9.5 billion in the essay. They hold limited rights inside vault contracts: which lending markets receive funds, which risk parameters are set, and how collateral is watched. Despite that sum, he says the revenue is very limited, especially next to traditional asset managers with more discretion and more responsibility. Many managers charge no fee at all. The largest DeFi risk curator by assets, Steakhouse Financial, charges a management fee of 22 basis points; the average across managers is 19.

Part of the reason, he suggests, is that risk-adjusted DeFi yields have not been attractive, possibly linked to idle capital in Aave. Another is structural: Morpho and Kamino have commoditized the vault layer and sit between the risk manager and the depositor—distribution again. The larger constraint is that the managers mostly use the same data. If the collateral is ETH or BTC, price, liquidation rules, and collateral information are public. AI makes a risk model easy to copy, so managers compete on judgment—risk appetite, much of it already defined by algorithm—and on a track record. Over a long enough span, he expects that to pressure fees.

OnRe is the contrast: an on-chain reinsurer that raises capital from investors and uses it to underwrite real-world insurance risk. That requires historical claims data, loss probabilities, the insurance contracts themselves, and people who understand them. None of that is available from a block explorer. A frontier model without access to the reinsurance data cannot do the underwriting, however strong it is at writing software. That, he writes, is why OnRe can charge 75 basis points plus 20 percent of profits. At the return rate then current, annualized fees were about 3 percent—16 times the average fee of DeFi risk managers.

He notes that these are different risk profiles, and that part of the premium reflects illiquid assets and the chance of loss. That is his point. A firm with proprietary data, underwriting skill, and client relationships can charge fees that a look-alike business cannot. If the data and the tools are the same as everyone else’s, the business is on the way to being a commodity.

Chains versus applications

A pessimistic reading he is willing to spell out: tokens tied to general-purpose L1s and L2s are in a difficult position, because a blockchain used to supply a new DeFi application with two of the three moats, distribution and capital.

Distribution, historically, came from users leaving assets on the chain where they were most active. Build on Ethereum and you reached Ethereum users; build on Solana and you reached Solana users. Starting Uniswap, Compound, Aave, Kamino, or MakerDAO from nothing, that mattered. There were no existing customers, assets, or financial relationships, so launching where users had already gathered made the start easier. Interoperability, he argues, is wearing the advantage down. Users can enter through an application and trade assets on almost any chain without thinking about the chain underneath. More new users, he expects, will arrive through fintechs that already have tens of millions of customers—Revolut, Nubank, and Robinhood—without first funding a wallet.

L1s also supplied capital through an ecosystem wealth effect. As SOL and ETH rose, users had more to put into applications on those chains. Difficult bridging, and a habit of loyalty to one chain, made capital more likely to stay. Bridging is easier now, and capital can enter as stablecoins, tokenized assets, and institutional money, so a new protocol depends less on any single token’s wealth effect. Issuers of real-world assets want assets on every chain; he names Securitize, Superstate, and Ondo as already showing that pattern.

On data, he says blockchains never had much of a moat. Almost all on-chain financial information is public, so neither the chain nor the application held much proprietary data that could serve as a barrier. AI, he adds, also weakens an older advantage: the developer ecosystem. Ethereum’s edge over newer chains such as Solana or Aptos included mature tools, audited code, and a dense set of integrations. As AI makes it easier to move an application across virtual machines, that edge matters less.

General-purpose blockchains, on this account, are losing their distribution and capital advantages and never really built a data moat. They compete mainly on execution and settlement, where blockspace is plentiful and fees keep falling. He does not conclude that a blockchain is therefore worthless. He points to a summit talk by Tushar a few years earlier, in which an L1 token is described as having features of a capital asset, a commodity, and a store of value. The illustration: if Solana eventually processed a billion payment transactions a day and charged $0.01 each, annual revenue would be $3.65 billion. That is a hypothetical, not a reported run-rate. A Blockworks Research chart in the essay is described only as showing that volume was still growing; the text quotes no series.

The core conclusion he draws is that the source of value for L1s and L2s may be entirely different from what investors first expected.

For applications, the businesses he expects to capture the most need at least one of three things:

  • Distribution: a direct customer relationship that is hard to leave. His examples are Robinhood and Hyperliquid.
  • Capital: a balance sheet that can absorb losses.
  • Information: the ability to underwrite using data other participants cannot get.

However strong the models become, he writes, those three remain far harder to copy than software.

He says it does not much matter to him whether value accrues to equity or to a token. Both can capture it. Recent SEC guidance and proposed rules have, in the essay’s wording, opened a path for tokens to capture value in a compliant way, and Multicoin “fully plans” to invest in opportunities of that kind. That is a manager’s stated intention. It is not evidence that a token will produce a return. What he says matters is whether the business—token or equity—controls something scarce, whether that confers pricing power, and whether the advantage can be turned into a return for investors. As one of the larger investors in the field, the firm says it spends its time on which projects across the stack actually have these moats and can keep them. As more financial activity moves on-chain, he expects the gap between a protocol with a moat and a protocol that only has software to widen.

A wedge he refuses to call a moat

The easy conclusion, which he calls reasonable, is that incumbents win. Robinhood has the customers. Goldman Sachs may have more capital than any DeFi prime broker. Moody’s has the data and the licenses. If building software is becoming free, and the real moats are expensive resources, the firms that already hold those resources have the advantage.

Startups, he says, still have one edge: a willingness to take risks incumbents will not. AI cuts both ways. The same tools that could let Robinhood ship a DEX in a weekend could let a three-person company do it. As the cost of a competitive product falls, the opening is markets the giants will not enter and customers they do not serve.

His examples are specific businesses, not a sector basket:

  • Ethena built a synthetic dollar around a delta-neutral hedge, the basis trade. Traditional finance knew the strategy and, he says, previously could not or would not bring it onto crypto exchanges. Assets had reached $5.4 billion.
  • Pump.fun and fomo built new trading experiences around memecoins and, from that entry, large distribution channels. He calls them among the most successful projects of this market cycle. That is his judgment about the cycle’s product winners, not a full account of revenue or of how long users stayed.
  • Hyperliquid entered a derivatives market dominated by Binance, Bybit, and OKX and took share quickly. It also began listing perpetuals on pre-listing companies, including OpenAI and Anthropic, markets he says traditional exchanges will not touch. Those markets, he writes, drew new traders and liquidity. Wallet creation on the day of the SpaceX IPO is his illustration, cited to HyperTracker; the essay does not give the count.
  • On capital, Project 0 finances DeFi collateral that traditional prime brokers will not accept.
  • On proprietary data, USD.ai underwrites loans against GPUs, collateral with which banks have almost no experience.

The common thread he draws is entry into markets incumbents were not serving—because of regulation, operational complexity, reputational risk, or an opportunity that looked too small. As long as that holds, a startup can enter asymmetrically. Crypto’s history, he says, has shown that more than once, and he expects it to continue. AI lowers the cost of the financial product; DeFi lets a startup reach customers and capital without rebuilding an entire financial system.

He is explicit that risk tolerance is not itself a durable moat. Regulation catches up, and incumbents enter once a market is validated. Spot bitcoin is his example. A startup does not need the moat on the first day. It needs a wedge the incumbent will not enter, and enough time to turn that wedge into one or more of the three moats, and then to compound it.

What the essay does and does not show

Applebaum’s close restates the decade-scale bet. Trillions of dollars of financial activity move on-chain, and AI makes the products that serve it easier to build than before. Software is being commoditized. Customers, capital, and proprietary data are still hard to obtain, and the firms that hold them have pricing power. That is why Multicoin says it is constructive on the next generation of DeFi companies. Code can be copied, or regenerated with the latest model. The moats, he writes, cannot be generated directly, and that is the opportunity he sees as a crypto investor.

Read against the categories the essay itself separates, a general-purpose chain, a wallet, a perpetual-futures venue, a stablecoin issuer, a credit curator using public prices, and a reinsurer using claims files are not the same business and not the same exposure. Attention—a memecoin interface, a new perpetual market—is not the same fact as funding, and neither is the same as adoption that is costly to leave: funded accounts, a balance sheet built from retained earnings, or data no one else can query. The essay marks several limits itself. User-owned liquidity can leave. Incentivized liquidity often leaves when the incentive ends. Frontends fall into price wars. Chain-level distribution is eroding as interoperability and fintech onboarding improve. A willingness to take risks incumbents avoid is temporary once the market is proven. Where the essay cites a chart for DEX share, tokenized assets, or trading volume, the text does not supply the underlying figures, so those claims stay at the level of the author’s description.

Nothing here is a recommendation to buy a chain, an application token, or a basket of “narrative” assets, and an isolated essay is not evidence of a rotation of capital.

Source: panewslab.com (originally published 2026-11-11 10:29 UTC+8)

Further reading