On November 11, 2026, at 11:22:56 a.m. China Standard Time (UTC+8), TechFlow published a Chinese compilation of an essay by Spencer Applebaum of Multicoin Capital. The byline credits Applebaum as author and TechFlow (深潮) as compiler. The piece argues where pricing power sits in DeFi once software is cheap. It is not a product announcement. Figures below are the essay’s, with the sources it names. It points readers to a disclosure on the Multicoin original; that disclosure is not reproduced in the TechFlow text, and no separate Multicoin URL is given here. Multicoin describes itself in the essay as one of the largest investors in the sector. That stake is part of how its conclusions should be read.
What the essay says got cheap
Applebaum’s comparison is concrete. Five years before the essay, building a DEX or a lending protocol, in his account, meant raising $3 million, hiring five engineers, and spending six months on a minimum viable product. He argues that a founder using Claude Code can now put up most of the frame in a weekend, and that a frontier model asked to build a perpetual-futures exchange can generate a matching engine, a risk engine, a trading interface, an API, and most other core components nearly out of the box.
The claim is not that finance becomes a commodity. Customer lists, a large balance sheet, and proprietary data plus underwriting skill take years of relationships, a track record, and trust. A model, he writes, can produce a lending protocol’s code in an afternoon. It cannot produce $10 billion of stablecoin deposits flowing into that contract, and it cannot bring 28 million customers to a new internet brokerage. Those two figures are illustrations of what software cannot mint. Later, and separately, the essay says Robinhood has 28 million funded accounts.
The premise is stated as a belief, not a completed result: DeFi continues to take activity from centralized crypto finance and, in time, from traditional finance. The data point offered for the first half is DEX spot volume as a share of CEX volume. The essay says that share was at a historical high and still rising, and that it had recently reached 24 percent, citing The Block and DefiLlama. The text refers to a chart of the ratio. It does not print the rest of the series.
On tokenized assets, the essay says U.S. Treasuries, private credit, and equities are seeing real on-chain use, and that the growth is no longer only from crypto-native users. Traditional firms, fintech companies, institutional asset managers, and traditional brokerages are described as building on-chain products and bringing existing customers with them. A chart of tokenized real-world-asset scale and mix is attributed to rwa.xyz. The prose states no aggregate dollar total, and none is added here.
The strategic conclusion is that today’s large DeFi protocols do not automatically collect that growth. As the cost of software and of data analysis falls, value moves toward whoever controls inputs software cannot create. The essay names three moats: distribution, capital, and proprietary data. A fourth factor, willingness to take risk, is treated differently — a wedge into a market, discussed at the end, not a moat of the same kind.
This is presented as the next pass in a longer Multicoin argument. Earlier versions looked at application tokens capturing value through work tokens and burn-and-mint equilibrium; then at “unforkable state,” on the view that anyone can fork open-source code but not liquidity or network effects; then at the idea that protocols capture the most value when they must manage real risk. The essay says those points were partly borne out.
Distribution: who can send the order, and who can move the balance
Distribution, as the essay defines it, is the ability to decide where order flow or fund balances go. Not every frontend is treated as equal. An identified, funded account that is hard to leave can be a real barrier. A pure interface is easier to contest.
The stablecoin stack is the clearest pricing-power example in the piece. Issuing a stablecoin at scale, when rates are not near zero, is described as a good business: Tether and Circle hold dollar reserves and earn the interest, and in most cases share little or none of it with holders. Since 2021, the essay says, Tether alone has earned more than $20 billion of interest income at a margin close to 100 percent. A DefiLlama series it cites runs from May 2021 through October 2026 and puts cumulative revenue at $20.2 billion.
That income is issuer economics. On the essay’s own description, holders are generally not the ones receiving it. A cumulative interest figure also does not establish that reserves match liabilities holder by holder, that the reserve assets are of any stated quality, or that the total is an audit. A stablecoin balance visible on-chain is a token position. It is not, by itself, legal title to the reserve portfolio, and it is not evidence that the issuer or a distributor is solvent.
The essay then turns the same chain around at the venue. It says more than $6 billion of USDC sits on Hyperliquid as derivatives margin, and that those retained balances create reserve interest for Circle and its distribution partners. Hyperliquid’s bargaining point, in this account, is that the balances are there because of the venue. If Circle would not share the yield, Hyperliquid could switch to a stablecoin that would, or in principle issue its own.
Earlier in the year of publication, the essay says, Hyperliquid reached an agreement with Coinbase under a framework the text calls aligned quote assets (AQA). Under that agreement, about 90 percent of the USDC reserve yield on the platform no longer accrues to Circle or Coinbase. It goes to the protocol’s Assistance Fund and is used to repurchase HYPE. Taking the $5.5 billion of USDC that the essay identifies as that 90 percent portion in the vault, the arrangement is said to be worth about $185 million a year to Hyperliquid. The essay stresses what Hyperliquid does not have to run: mint-and-redeem pipes, treasury management, or a compliance license. Control of the inlet is treated as enough to demand a larger share of the economics.
Read as a map of control and obligation, these are different roles. Users post USDC as margin; that collateral is theirs to lose in trading, not Hyperliquid’s equity. Circle is the issuer associated with the reserve yield. Coinbase is the distribution counterparty in the reported agreement. Hyperliquid controls the venue where the balances are in use. The Assistance Fund is where the essay says the redirected yield goes, and the use it names is a HYPE buyback. A contractual split of yield is a commercial term. It does not show who would be paid, or in what order, if redemptions, liquidations, and a halt arrived together. It does not turn a token buyback into customer protection.
Wallets charge for the relationship the venue still has to earn
The other side of the argument is wallets and frontends. Early non-custodial wallets, the essay says, struggled commercially, then recognized that trading was the high-margin business and made it a core product. On swaps, MetaMask is said to charge up to 0.875 percent. Uniswap’s average trading fee over the 12 months before the essay is given as 0.13 percent, with most of that paid to liquidity providers rather than to the Uniswap protocol. In this comparison the wallet earns more than the place that matches the trade. The venue has to attract liquidity providers, maintain infrastructure, and compete on slippage. MetaMask has to win and keep the customer, and can route the order elsewhere if a quote is poor. Over time, the essay says, a frontend with enough flow can internalize a large share of trading and keep more of the economics.
A similar split is described in perpetuals. Phantom lets users trade Hyperliquid perps and adds 5 basis points on top of Hyperliquid’s native 4.5 basis points. Hyperliquid runs the risk engine, the matching, and the liquidity. Phantom, holding the interface and the user, takes the higher cut. The essay also marks a limit. In the Hyperliquid Builders field, which it calls a newly contested market, a price war is underway and teams are cutting fees to retain users. The frontend layer is not immune to being commoditized.
Crypto wallets are called relatively fragile because the user’s cost of leaving is low. Import a private key or a seed phrase, and the same assets are available in another interface within minutes. Replacing a bank or a fintech app is harder: a new account, KYC, a transfer over traditional clearing rails, a re-linked payroll deposit, and updates to connected services.
The funded account is the scarce asset
Robinhood is the example of distribution that is harder to leave: 28 million funded accounts. Uniswap is described as still the main DEX on Robinhood Chain, an arrangement the essay says Robinhood accepts. The essay does not claim Robinhood will build its own DEX. It notes that sitting as principal on liquidity has not historically been Robinhood’s core aim, while also noting that Robinhood has begun sending a large volume of prediction-market orders to its own venue, Rothera, rather than to Kalshi. The narrower point is that writing the protocol is no longer the scarce step. Users, capital, and proprietary data are.
The picture that follows is a waterfall. Frontends such as Phantom take a larger share from the venue. The venue, Hyperliquid, takes yield back from the stablecoin issuer, Circle. At each step, the essay says, the party with distribution has pricing power over the infrastructure below. How much profit stays at a layer depends on how hard it is for the user to leave. Hyperliquid is described as keeping about 90 percent of the reserve yield, while frontends running on top of it are already being pushed to cut add-on fees to only a few basis points.
Capital: software does not post margin, and it does not take the loss
The easy inference — that a protocol with no frontend is simply stuck — is only partly accepted. Capital is the next moat. Traders need it to use leverage, market makers need it to keep quoting, and lending protocols need deposits before they can lend. A model can write code and automate a workflow. It cannot create the funds. When a system breaks, someone has to pay. The essay names Aave’s KelpDAO episode and MakerDAO’s Black Thursday as cases of that fact, without further detail.
Liquidity is treated as real but not the same as a durable capital moat. The network effect is described as logarithmic: traders go where liquidity is, and liquidity follows traders. Hyperliquid’s open interest is called an advantage, but that money belongs to users who can withdraw it. The essay recalls liquidity moving among DeFi protocols under token incentives and vampire attacks, then leaving when the incentives fade. Two other forms of capital are what it treats as longer-lived defenses.
One pool of margin across venues that do not share a book
Prime brokers in traditional markets provide leverage and financing, and they let a hedge fund use one pool of capital across markets instead of locking margin at every venue. The essay calls the business very large. Industry prime-brokerage revenue is said to have grown from about $15 billion in 2020 to about $27 billion in 2024, with the hedge-fund industry and financing-heavy multi-strategy “pod” firms.
The DeFi version, in this argument, could be larger. Traditional prime brokers mainly serve large institutions because small clients are expensive to operate. Contracts can automate most of the process, so the essay’s range runs from a retail trade of $500 to a Millennium research pod running positions in the hundreds of millions of dollars. Today, it says, a trader typically posts separate margin on Hyperliquid, Aave, Lighter, Morpho, Variational, and Kamino. That uses capital poorly, and the problem is expected to worsen as tokenized equities, Treasuries, and private credit move on-chain. A fund that is long a tokenized stock in one venue and short an equity-index perpetual in another should not, where some of the risk offsets, have to post full margin on both sides.
Software that tracks those offsets is getting easier to build. What remains hard is a balance sheet large enough to finance the positions, a low cost of funds, and risk experience that avoids losses beyond recovery. A prime broker with cheaper funding can offer better terms, attract more clients, and then fund itself still more cheaply. That is the scale loop described. A large stablecoin balance at a venue is not that balance sheet. It is not the venue’s capital, and it does not say who is obliged to cover a shortfall.
The money that stays when a higher yield appears somewhere else
The second use of capital is standing behind protocol risk. This continues an earlier Multicoin argument: a protocol token that does not itself manage risk is easier for a competitor to fork and replace than one attached to a protocol that does.
Hyperliquid’s HLP vault is the example. It supplies liquidity by making markets and by taking on liquidated positions. Depositors share the profits and also bear the losses. In March 2025, the essay says, HLP lost about $4 million handling the liquidation of a large ETH position. A later JELLYJELLY episode put the exposure under pressure again. Both are cited as reasons a derivatives DEX needs something in the role of an insurance fund. Aave’s Umbrella is placed in the same family: users take bad-debt risk in exchange for extra yield.
Anyone can fork the HLP or Umbrella contracts. A fork does not take the money locked inside. The caveat is that this money is third-party funds. If a higher yield shows up elsewhere, depositors can leave. Some stickiness is attributed to overlap between token holders and vault depositors: a HYPE holder might deliberately leave dollars in HLP to support the exchange and to protect a token exposure. The form the essay calls most durable is a balance sheet the protocol builds from retained earnings — capital that does not leave because another project offers a higher short-term yield or a points campaign.
Binance is offered as evidence of what an owned balance sheet can do. After a $6.3 billion theft on October 29, described as recent in this November 2026 essay, Binance is said to have filled the hole in full with SAFU reserves and company funds, mainly the latter, resumed withdrawals within a few days, and seen net assets leave by only 3 percent. The essay says this was still early, but that confidence in the size of the balance sheet clearly helped stop a run.
That report has a narrow meaning. Covering a stated loss with SAFU reserves and corporate money is a counterparty choice, followed by a reported net-outflow figure. It does not establish the exchange’s full liabilities, the priority of customer claims, or an audit. It does not follow from any wallet balance. The essay’s own distinction is the one that matters: capital whose owners can withdraw it is not the same obligation as capital the firm has retained and is willing to lose.
As DeFi moves further into private credit, tokenized equities, and other real-world assets, the essay expects demand for insurance capital and risk capital to keep rising. Protocols that supply and run that capital, it argues, will capture more value than parties that only ship software.
Proprietary data: a public ledger prices public risk
The third moat is information, with an explicit condition. DeFi, the essay says, was built on the idea that financial data should be open, and that idea works for overcollateralized crypto lending and for spot trading. As more real-world assets move on-chain, models built on non-public information become more common. The moat described is the combination of proprietary non-public data and the ability to price and underwrite from it — not data as a slogan.
Lending shows the split first. On Morpho, Kamino, and Euler, risk curators managing funds for depositors are said to oversee $9.5 billion, cited to DefiLlama. They receive limited permissions inside vault contracts: which lending markets to allocate to, which risk parameters to set, and how to monitor collateral. Despite that $9.5 billion, the essay says the revenue they earn is small, especially next to traditional managers with wider discretion and responsibility. Many charge almost nothing. Steakhouse Financial, identified as the largest DeFi curator by assets, is given a blended take of 22 basis points. The average take across curators is 19 basis points.
One reason offered is that risk-adjusted DeFi yields have not been especially attractive, perhaps partly because a large amount of idle funds sits in Aave. The “perhaps” is the essay’s. Another reason repeats the distribution point: Morpho and Kamino commoditized the vault layer and sit between curator and depositor. The deeper reason is that most curators use the same public data. For lending against ETH or BTC, price, liquidation rules, and the collateral are visible to everyone. AI makes a risk model easy to copy, so the essay expects curators to be pushed toward one-off judgment and a track record, and fees to stay under pressure over time.
OnRe is the contrast. It is described as an on-chain reinsurer that raises capital from investors and uses it to underwrite real-world insurance risk. Doing that requires claims history, loss probabilities, and the text of the insurance contracts, plus actuaries who can read them. None of that is available in a block explorer. A frontier model, however capable, cannot price the cover without the internal reinsurance book. That is why, in this account, OnRe can charge a management fee of 75 basis points plus 20 percent of excess returns. At the return level then current, the annualized fee rate is put at about 3 percent, roughly 16 times the average DeFi curator.
The essay does not treat these as the same business. It says the fee premium also compensates for illiquidity and for the risk of losing principal. The comparison is used for a narrower claim: a firm with proprietary data, underwriting skill, and a client relationship can set a much higher price; a firm with the same public data and the same tools as everyone else tends to end as a cheap commodity.
Assets visible inside a curator vault are not the same thing as legal ownership of those assets. A curator’s limited rights inside a contract are not a promise that depositors are repaid. A published fee does not say whether the underlying positions can be exited, or who bears the loss if collateral fails.
Chains lose an old subsidy. Applications still need one scarce input.
Applied to base layers, the reading is deliberately cold. L1 and L2 infrastructure tokens are in an uncomfortable position, the essay argues, because a chain used to hand new DeFi applications two of the moats — distribution and capital — and that subsidy is fading.
Historically, a chain had distribution because users kept assets on the most active chain. Building on Ethereum reached Ethereum users; building on Solana reached Solana users. When Uniswap, Compound, Aave, Kamino, or MakerDAO started, launching where users already were mattered. Cross-chain interoperability is described as weakening that bond. A user of the app fomo, the essay says, can open an account in a phone app and trade assets across chains without caring which chain is underneath. More new users, in this forecast, will arrive through Revolut, Nubank, and Robinhood, firms that already have tens of millions of users and do not require those customers to manage their own wallets.
Chains also used to supply capital through a wealth effect. After SOL or ETH rose, users had more to deploy on that chain, and bridging friction plus community loyalty kept it there. Bridging is now described as smoother. New money arrives through stablecoins, tokenized assets, and institutions. Issuers including Securitize, Superstate, and Ondo are said to prefer deploying on each major chain rather than on one.
On data, the essay says a general-purpose chain almost never built a moat. Most on-chain financial information is public by design, so neither the chain nor an ordinary application starts with defensible private data. AI is also described as eroding another old advantage, the developer ecosystem. One reason Ethereum led newer chains such as Solana or Aptos was mature tooling, codebases that had been audited repeatedly, and a wide set of integrations. If a model can rebuild an application across virtual machines, that advantage matters less.
The result, in this view, is that general-purpose chains are losing the old distribution and capital dividends and did not have a data barrier. They compete on execution and settlement. With blockspace plentiful, the fee per transaction is still falling. The essay does not conclude that a chain therefore cannot be valuable. It points to an earlier summit argument by Tushar — no surname is given in the text — that an L1 token combines a capital asset, a commodity asset, and a store of value. The arithmetic that follows is a hypothetical, not a reported run-rate: at one cent a transaction, if Solana eventually processed one billion payments a day, annual fees would be $3.65 billion. Separately, Blockworks Research is cited for Solana reaching 162 million transactions a day in 2026. Those are different figures. The lesson drawn is that the reason a chain is valuable may not match what investors first assumed.
For applications and protocols, the essay says the winners that capture the most value will hold at least one of the three moats: distribution through a direct customer relationship that is hard to leave, with Robinhood and Hyperliquid as the examples; a balance sheet that can absorb losses; or proprietary data with underwriting others cannot readily match. All three are described as harder to copy than software, whatever the model can do.
Whether value ends up in company equity or in a protocol token is treated as not decisive. Both, the essay says, can share in the shift. It says recent SEC guidance and draft rules have outlined a path for tokens to capture value in compliance with those rules. The text does not specify the guidance, and this account does not infer it. Multicoin says it will invest in that opportunity. That is the firm’s stated posture. It is not a finding that any token or share is sound, and it is not a recommendation.
The test the essay wants used, whether the wrapper is a token or equity, is whether the business controls a scarce input, whether that control gives it pricing power, and whether the advantage can be turned into a return for the investor. As more financial activity moves on-chain, the gap in valuation and earnings between protocols with those moats and protocols that are only software is expected to widen.
Incumbents already hold the moats. The opening is what they will not underwrite.
The objection is stated plainly. If software is close to free and moats are expensive, incumbents should be favored. Robinhood has the customers. Goldman Sachs, in the essay’s view, has a balance sheet larger than any DeFi prime broker’s. Moody’s has deep data and an authoritative license. Firms that already hold the moats should have the head start.
The card left to startups is a willingness to take risks those firms will not. AI cuts both ways. The same tools that could let Robinhood stand up a DEX over a weekend give a three-person team the same speed. As the cost of a working product falls, the opening is a market or a customer an incumbent leaves alone because of regulation, operational complexity, reputation, or because the niche looks too small at the start.
The cases, with the scale the essay states, are these:
- Ethena built a synthetic dollar around delta-neutral hedging. Basis trading is not obscure on Wall Street, but traditional firms, for compliance or internal reasons, could not or would not run it on crypto exchanges. The essay says Ethena reached $5.4 billion of assets under management by taking that gap.
- Pump.fun and fomo started near zero, built trading experiences around memecoins as a new asset category, and captured large distribution. The essay calls them two of the standout winners of this crypto cycle.
- Hyperliquid entered perpetual futures then dominated by Binance, Bybit, and OKX and took share. It also listed pre-IPO perpetuals tied to private companies including OpenAI and Anthropic, products the essay says a traditionally licensed exchange would not touch. That kind of market, it argues, keeps drawing new traders and added liquidity. HyperTracker is cited for daily new wallets on Hyperliquid from May 16 to June 14, 2026, with new wallets exceeding 40,000 on the SpaceX IPO day.
- On capital, Project 0 is described as financing against DeFi collateral that traditional prime brokers will not take.
- On data, USD.ai is described as able to underwrite and price loans secured by GPU hardware, a collateral type on which traditional banks are said to have almost no experience.
The common thread is that each entered ground left open by regulatory constraint, operating complexity, reputational concern, or small initial size. The essay treats that pattern as likely to recur. What it finds consequential about AI meeting DeFi is specific: AI cuts the cost of building the financial product, and DeFi lets a small team reach global customers and move capital without rebuilding an entire traditional-finance stack.
Risk tolerance alone is not called a lasting moat. Regulation catches up, and once a market is proven, incumbents enter. Spot Bitcoin ETFs are the precedent named. A startup does not need a finished moat on the first day. It needs an opening the incumbent will not take, and time to turn that opening into one or more of distribution, owned risk capital, or proprietary data. Only then does the essay expect the advantage to compound.
What remains scarce, and what a balance still does not prove
The close is conviction, not a measured flow of funds. Over the following decade, the essay says, trillions of dollars of financial activity move on-chain, and AI takes the cost of the software serving that activity to a low not seen before. Direct customer relationships, real risk capital, and non-public proprietary data do not cheapen with the code. Businesses that hold those inputs keep the ability to set a price. Code can be forked, or written by a leading model. No model generates the moat. The excess-return opportunity the authors point to, as investors, sits there.
Two distinctions follow from the evidence the essay actually assembles.
On-chain visibility is not legal ownership. USDC posted as margin, assets sitting in a curator vault, and tokens held in a wallet can all be counted. Counting them does not identify who has title to underlying reserves, who is owed a redemption, or who has agreed to take a loss.
A reserve total, a yield split, or a rapid refill is not a matching of liabilities, and it is not solvency. Tether’s interest income, the Assistance Fund’s reported share of USDC yield, depositor capital in HLP, and Binance’s use of SAFU plus corporate funds answer different questions. None of them, as reported in this essay, is an audit of what is owed to customers. A wallet balance does not establish one.
The TechFlow article, published November 11, 2026, is the source for this account: https://www.techflowpost.com/zh-CN/article/35279.