The Custody Arbitrage: Binance, Coinbase, Revolut and Nubank Are All Building the Same Onchain Product

Coinbase's last public count was north of 100 million verified users. Base — the rollup Coinbase built and controls — does not have 100 million people doing anything on it. Not close. Not within an order of magnitude. Sit with that gap for a second. It is the most valuable unpriced spread in crypto right now, and it explains why inside a single news cycle Binance, Coinbase, Revolut and Nubank all pointed their product roadmaps at the same three syllables: on-chain. Four companies. Four regulatory addresses. Four completely different customer bases. One identical conclusion — that the next ten million crypto users will not download MetaMask, will not write twelve words on a piece of paper, and will never knowingly touch a bridge. They will tap a button inside an app they already trust, and something will happen on a blockchain they will never be able to name. Whoever owns that button owns the fee. That is the whole trade. Everything else in the announcement cycle is decoration. I have been on both sides of this arbitrage. In the 2017 ether rush I was chasing the white whale — manually scraping forty-odd whitepapers off contract deployments, publishing buy/sell/pass calls on Telegram before the exchanges had finished their listing notices. Back then the edge was speed of information. In 2020 I audited Uniswap v2 and Compound contracts during DeFi Summer and found a slippage path in an early yield aggregator, ran it once for about twelve thousand dollars off student loan savings, then wrote the post-mortem. The edge there was speed of execution. The edge in 2026 is different, and it is the one nobody is writing about. It is distribution. The chains already work. The contracts already work. What does not work is the last mile between a person with money and a wallet they can operate without fear. Four of the largest financial interfaces on earth have just decided to fix that last mile themselves, and in doing so they will quietly become the most important layer in the stack — more important than the L1s, more important than the rollups, more important than the protocols whose names are on the tokens. Here is the confirmed part, stripped of varnish. Binance, Coinbase, Revolut and Nubank are all rolling out new onchain products. That is the fact. The details are thin — no technical specs, no token models, no timelines anyone will sign their name to. Which means the reporting you are about to read elsewhere will be a summary of a trend. What follows is a structure. Start with the market context, because it determines the tone of everything. We are in chop. Sideways. Days that open green and close red and give back the difference by Thursday. Chop is not a dead market; chop is a positioning market. Price is going nowhere, so narrative has to do the work. When an asset can't appreciate, product teams get loud. Every announcement cycle you have ever seen that felt premature was priced by someone who understood that flat price is exactly when the story has to carry the weight. So the timing here is not an accident. It is a schedule. Now the four players, because their onchain products will not look alike. They will look like their balance sheets. Binance is the volume incumbent with a listing monopoly it has been slowly losing to onchain venues for four years. Its problem is defensive. Every ticker that goes live on a DEX first — and stays there — is a ticker Binance no longer controls the price discovery for. An onchain product for Binance isn't expansion; it's reclamation. Watch for anything that funnels user flow through BNB Chain infrastructure, because that is where Binance can tax the traffic without a regulator being able to point at a securities offering. Coinbase is the compliance-first, US-listed venue. Its onchain product has a different job: it needs to convert a custodial user base into onchain activity to defend a valuation multiple that assumes an internet-scale platform rather than a brokerage. Coinbase has spent years building the compliant rail — the licensed derivatives venue, the custody arm, the rollup. The onchain product is the payoff leg. And here is the tell that matters most: Coinbase will not ship anything that cannot survive an SEC examination, which means its onchain product will ship with a door that closes. Quietly. Legally. Revolut is the FX super-app, roughly fifty million retail customers across the UK and EU, crypto already a live revenue line, and a regulatory posture that makes it permanently cautious in public and aggressive in product. Revolut's edge is not blockchain. Revolut's edge is that its customers already move money across borders every day and already hate the spread they pay for it. Stablecoin rails on top of an FX book is one of the very few places in this industry where onchain settlement is genuinely, unarguably cheaper than the incumbent alternative. Not narrative-cheaper. Actually cheaper. Nubank is the one I would watch closest, and almost nobody in the anglophone crypto press is. A hundred million customers, mostly Brazil and wider Latin America, in economies where dollar access is a practical need rather than a portfolio allocation. Nubank has already made public moves toward holding bitcoin on its own balance sheet as a reserve asset. That is the signal. A neobank holding BTC as a reserve and building onchain product on top of a hundred million retail accounts in a high-inflation corridor is not experimenting. It is executing. Four different problems. One answer, because the answer is cheap now and expensive later. Here is the plumbing, and this is where the technical reporting is failing you. The unsolved problem in consumer crypto was never consensus. It was key management and gas. Twelve words written on paper, one seed phrase in a screenshot folder, one wrong signature and everything is gone forever — that is the failure mode that kept a hundred million Coinbase users custodial. Not ideology. Loss aversion. People do not self-custody because self-custody is a skill, and skills have a failure rate. Account abstraction solved the mechanics. Smart accounts, passkeys, session keys, paymasters that sponsor gas so the user never has to hold a native token to make a transaction. Social recovery that replaces the paper phrase with three trusted contacts. This stack is not exotic anymore. It has been in production on multiple chains for years, and it works well enough that the remaining barrier is not engineering but integration. That integration is what four companies with existing app installs can do and what a standalone wallet cannot. Binance does not need to teach anyone what a seed phrase is. Revolut already has a payment card in the user's phone. Nubank already has the user's paycheck landing in an account every month. The distribution is solved. The compliance is solved. The only question left is the custody model, and the answer to that question decides whether any of this is real. So let's talk about where the money actually is. Trader's lens, because the strategy decks won't show you this. A custodial venue that hosts your onchain activity captures the spread, the withdrawal fee, and — critically — the float. Not the protocol fee. The venue fee. When a protocol earns five basis points on a swap and the interface in front of it earns thirty, the interface is the business. This is not a new observation. It is the observation that every DEX aggregator has built its margin on top of, and it is why the largest wallets in the industry make more revenue than the protocols they route to. Now multiply that by four companies with hundreds of millions of combined retail accounts. The reason none of them will launch a token is that they do not need one. A token would dilute the exact revenue line they are trying to grow. Their equity, or their stablecoin float, or their reserve asset is the value capture vehicle. That is the silence in the tokenomics section of every one of these announcements, and it is not an oversight. I learned this the hard way in the 2021 minting frenzy, minting ghosts at light speed — a hundred and fifty units across early Punk and Ape derivatives, tracking gas wars on Etherscan to see how congestion killed mint success rates. The lesson was not about NFTs. The lesson was that the platform selling the mint mattered more than the art it minted. The same physics apply here, one layer up. There is a parallel in Bitcoin that almost nobody wants to say out loud, so I will. After the fourth halving, block subsidy became a rounding error against fee revenue for a lot of the mining sector, and miner revenue compressed hard enough to shake out the marginal hashrate. What survived consolidated. Publicly, we still describe Bitcoin as the most decentralized network in existence. Operationally, the block template is produced by a shrinking set of pools, and the number that actually matters — how many independent entities could censor a transaction if they chose to — has been falling even as the hashrate chart goes to the upper right. I am not arguing Bitcoin is broken. I am pointing out that decentralization is a spectrum that markets price at zero right up until they price it at everything. Custody on the consumer side is going through the identical consolidation, just faster and with better marketing. The chain stays decentralized. The door to it gets a logo. And there is a second parallel, in the sector that has burned more investor capital on storytelling than any other: real-world assets. I have been watching tokenized treasuries, tokenized funds, tokenized private credit for three years. I have audited the revenue splits on adjacent infrastructure and I will tell you the uncomfortable thing. Institutions do not need your public chain. They need a settlement layer that satisfies their auditor, their regulator and their prime broker. Every time a serious institution has been given the choice between a permissionless network with pseudonymous validators and a permissioned environment they control, they have taken the permissioned one and called it onchain. The narrative said trillions were coming into public DeFi. What actually came in was a couple of custodial wrappers on top of money market funds, and a lot of conference panels. So when four regulated intermediaries announce onchain products, ask the uncomfortable question. Onchain how? Permissionless interaction, or a permissioned corridor with an onchain label on the marketing page? The answer will be in the withdrawal path. If you can send to an arbitrary address without a hold period, an allowlist, or a managed risk score, it is a real gateway. If you cannot, it is a fenced garden, and the fence is the product. I have watched enough cycles to know the fence usually wins. The fence has better UI. This is where the conventional read of the news gets it backwards, and I want to be precise about it. The standard take is that four giants adopting onchain products means adoption. More users onchain, more transactions, more TVL, more everything. Bullish for infrastructure, bullish for the chains these products deploy onto, bullish for DeFi protocols that get integrated as backends. Maybe. But that framing measures the wrong thing, and it has a blind spot the size of the entire consumer custody market. What is happening is not adoption. It is absorption. Adoption implies the user enters the ecosystem and becomes part of it — holds their own keys, interacts with protocols directly, becomes a participant in the permissionless layer. Absorption means the user never enters at all. They stay inside the app. The blockchain becomes a settlement backend, invisible, fully intermediated, technically present and functionally irrelevant to the person pressing the button. That distinction decides who captures the value of the next five years. If users become participants, the protocols win and the intermediaries are pipes. If users get absorbed, the intermediaries win and the protocols become commodity backends, competing on wholesale economics with zero brand loyalty and no pricing power. Look at what is actually being built and tell me which one is more likely. The entire engineering effort — the sponsoring of gas, the passkeys, the recovery flows, the hiding of addresses — exists to make the chain invisible. Nobody spends that much money making something invisible so their users can fall in love with it. The metric to watch is not TVL. TVL can be manufactured by a single custodial treasury and it will look identical on a dashboard to a thousand independent depositors. The metric to watch is the ratio of custodial-to-self-custody outflows. How many users who touch these new products ever move an asset to an address that the platform does not control? That number, published honestly, would tell you in one quarter whether this cycle is a gateway or a fence. Watch for it. It will not be published voluntarily. And a second-order effect nobody has priced. If these four succeed at making onchain activity frictionless inside a familiar app, the losers are not the DeFi protocols. The losers are the standalone wallets. The self-custody specialists. The companies that spent a decade teaching people to hold their own keys, and whose entire product is a slightly better version of that same friction. Those companies just got outflanked by a bank license and a payment card. Speed kills slower than greed. The venues that can move the fastest on distribution have every incentive to move slowly on disclosure. So what do you actually watch, starting this week, in a market that is going sideways and will not tell you where it is going next? Three things, and none of them are price. First, the smart account specification. The first of these four to publish an audited account abstraction implementation with a named relayer and a documented gas sponsorship model is the one that is serious. Everything else is a landing page. Second, the exit. Test the withdrawal path. Can you move assets to a non-platform address, at any hour, without a risk score gating you? How long does it take? Is there a maximum? The exit is the entire constitutional document of an onchain product, and it is written in the terms of service, not the whitepaper. Third, the stablecoin float. If a neobank or an exchange issues or holds its own dollar instrument to settle onchain flows, that instrument becomes the real balance sheet of the product. Follow the float, not the fee. Volatility is just noise until it becomes signal. Right now the signal is not in the candles. It is in the plumbing — four companies deciding, in the same season, that the fastest path to owning the next hundred million onchain users is to make sure those users never find out they are onchain at all. Ask yourself one question and sit with it. When the button finally arrives, and it will arrive inside an app you already have, whose keys are behind it — and will you ever be told?