On 28 September 2026, Citi and Coinbase announced the expansion of their partnership into a two-sided product: Citi's institutional clients accept stablecoin payments through the bank's merchant rails, and Coinbase's payment customers receive what coverage describes as a deposit-style account, built on Citi's banking infrastructure, that automatically sweeps incoming dollars into stablecoins held in Coinbase custody. The swept balances carry a reward of 3.75% a year. Most coverage read the deal as Wall Street going onchain. The more instructive reading runs the other way: the product is a public confession about what a deposit is.
A deposit is not the customer's money. It is a claim on a bank — a foundational principle of English banking law since 1848 — that behaves as money only because institutional machinery makes the claim exchangeable at par, so effectively that nobody has had to look at what sits underneath. The new account prices, per annum and in public, the question the deposit was engineered not to raise: what exactly do you hold, and against whom. This essay strips the label.
The arrangement has two directions. In one, Citi's corporate clients accept stablecoins at checkout through the bank's merchant-processing platform, with Coinbase running the payment rails and converting the digital asset into fiat, which Citi settles as bank of record. In the other, Coinbase's payments customers get account-like functionality built on Citi's virtual-account infrastructure: they can receive, hold and send fiat, and incoming dollars are automatically converted into dollar stablecoins that Coinbase custodies. On those balances, the customer earns 3.75% a year (The Block; Yahoo Finance, 28 September 2026).
The phrase used in the reporting for the second leg was "deposit-style account" (Yahoo Finance, 28 September 2026). The suffix is the honest part. The account behaves like a deposit at every visible surface — a balance, a payment capability, a familiar interface — while, on the product's public description, the thing actually held stops being a deposit the moment the sweep executes. Nothing about this is concealed. It is simply not itemised, in exactly the way the deposit itself has never been itemised.
That makes the product a rare specimen: an instrument that performs, in software and at machine speed, the transformation that the word "deposit" has been quietly performing in language for close to two centuries. To see what the sweep does, you first have to see what the deposit already was.
In English law, the foundational principle was stated by the House of Lords in 1848. In Foley v Hill, Lord Cottenham held that money paid into a bank ceases to be the customer's money at all; it becomes "the money of the banker," who may do with it as he pleases, owing the customer only an equivalent sum on demand (Foley v Hill, 1848). The relationship is debtor and creditor. The bank is not a warehouse, not a custodian, not a trustee. The depositor does not retain title to the deposited cash; what he holds is a claim on the bank — unsecured in form, cushioned in practice by the priorities and protections that insurance and resolution law later built around it. Other jurisdictions reached the same destination by their own routes.
The vocabulary predates the law. "Deposit" descends from the goldsmiths' business of the seventeenth century, when coin was in fact deposited — bailed, in legal terms — and the receipt was a claim to specific property. The goldsmiths discovered lending, the specific property became a pool, the receipt became a promise, and the law eventually followed the economics. The word did not. Ordinary language still encodes the warehouse theory: money in the bank, as if it had a location and the location were yours.
Every subsequent layer of banking law is built on the 1848 answer, not the vocabulary. Capital requirements exist because depositors are creditors of a leveraged debtor. Deposit insurance exists because the claim can fail. Resolution regimes exist to decide what happens to the claim when it does. The entire prudential apparatus is a monument to the fact that the deposit is credit — constructed, with some care, so that nobody holding one ever has to think about it.
Why, then, does the claim spend like money? Because moneyness is not a property of the instrument. It is a property of the machinery around the instrument. The distinction deserves one precise sentence: a deposit is money in the monetary system — the most common form of it the public holds — but it is not central-bank money, and it is not the customer's cash held in segregation.
Four pieces do the most visible work; capital, supervision, resolution and the settlement architecture stand permanently behind them. Par convertibility: any deposit ordinarily exchanges one-for-one into cash and into any other bank's deposit, without a continuously quoted market price. Deposit insurance: below the coverage ceiling, the claim's credit risk is transferred to a public backstop, which is what makes par credible for the public. The lender of last resort: the central bank stands behind a solvent bank's liquidity, so that a timing mismatch cannot force a solvent debtor off par — liquidity support, not a solvency guarantee. And settlement finality: payments between deposit claims discharge obligations irrevocably, by legal designation rather than by anyone's ongoing consent.
The Bank for International Settlements calls the result the singleness of money — the condition in which a unit of the currency is worth the same regardless of which balance sheet it happens to sit on (Garratt & Shin, 2023). Singleness is an achievement, not a default. Where the machinery is absent, claims on different debtors trade at different prices; the nineteenth-century United States priced banknotes against the issuing bank, and stablecoins in secondary markets deviate from par whenever their issuer is doubted. A deposit is money to precisely the extent that institutions make examining it unnecessary. This is the same operation this library has traced elsewhere: money is not a thing but an agreement maintained by infrastructure (see The Phenomenon of Money).
The machinery is not free, and the depositor is the one paying for it. The payment is just never presented as a price.
Banks typically move deposit rates less than one-for-one with the policy rate, and the spread this opens widens as rates rise — the deposit franchise, in the literature's term (Drechsler, Savov & Schnabl, 2017). The spread is the market price of everything Section III described: par at will, payment services, insurance, the option to leave at any moment without a mark-to-market. It is real, it is large in aggregate, and it appears on no statement. The depositor experiences it as an absence — interest not received — rather than a fee. An absence is the one kind of price that produces no invoice and therefore no comparison.
This is the gap the new product line attacks. Not the deposit's safety, and not its convenience — its pricing opacity. A sweep account that pays 3.75% on balances converted out of deposit form is, functionally, a counter-quote against the deposit franchise. It says: here is the opportunity cost of leaving the balance in deposit form, expressed for the first time as a number with a sign in front of it.
A deposit is money to precisely the extent that institutions make examining it unnecessary.
Follow one incoming dollar through the account. It arrives as a payment into banking infrastructure and exists, briefly, as a deposit claim on a global systemically important bank — insured within limits, inside the resolution perimeter, backed by the full prudential stack. Then the sweep executes. The balance is converted into a dollar stablecoin held in the customer's name at Coinbase. The customer now holds an interest in a custodied crypto-asset. The stablecoin itself is a claim on its issuer. Behind that claim stands a reserve portfolio — confined by the GENIUS Act, the 2025 federal stablecoin statute, to short Treasuries, repo, insured deposits and similar assets (GENIUS Act, 2025) — and the questions that decide outcomes are who holds the claim to those assets, under which law, and what happens on issuer insolvency. Any deposits in that reserve are, in turn, claims on other banks.
One visible balance; on the public description, a chain of three or four distinct claims underneath it. Each hop changes the debtor, the governing law, the failure regime and the remedy. A deposit claim on the bank sits inside deposit insurance and bank resolution. A custodial claim on an exchange sits inside custody law and the custodian's own insolvency, where the quality of segregation decides everything. A holder's claim on a stablecoin issuer sits inside the statute's redemption and reserve rules, which are new and untested at scale. None of these is the same thing as the others, and none of them is money in the bank.
The 3.75% is an observable price attached to this transformation, paid to the customer by the party that benefits from it — the number that makes the difference between the two forms of balance economically legible for the first time. How much of it is claim pricing and how much customer acquisition, promotion or strategic positioning, the announcements do not say. Nor do they disclose exactly the part that would let an analyst score the instrument: which stablecoins, what segregation structure, and what the customer's position is in each insolvency along the chain. The product prices the demotion without publishing the ladder.
The regulatory geometry is the most instructive part. The Act prohibits permitted stablecoin issuers from paying holders interest or yield for holding the coin (GENIUS Act, 2025). It says nothing about third parties. A distributor — an exchange, a platform — can pay rewards on custodied balances out of its own economics; whether those economics are fed by reserve interest shared upstream, by payment margins or by marketing budget is precisely what the public documents do not establish. The banking industry's associations understood the asymmetry immediately and spent the following year arguing that the ban should extend to affiliates and distribution platforms; the market-structure legislation that might have carried the extension failed to advance in the Senate (Yahoo Finance, 28 September 2026). Days later, a member institution of that lobby became the settlement bank underneath a distributor-paid reward product.
Read as hypocrisy, this is boring and slightly wrong. Read as position-taking, it is exact. Citi stands to earn bank-of-record settlement income and merchant fees on flows it does not have to hold as deposits: to the extent a swept balance no longer sits on the bank's book as a deposit liability, the funding relationship — and whatever prudential and accounting treatment attaches to it — has moved off that book. The precise consequences depend on contractual and booking structure the announcements do not disclose. Coinbase acquires account-like distribution without a banking licence, paying 3.75% as what functions, economically, as a customer-acquisition cost; whether the reserve economics upstream fund it, the announcements do not say. The customer receives the number. What is consumed in the process is the thing neither party has to account for: the claim quality the customer used to hold without knowing its name. The house rule of this library asks who benefits from the gap between label and reality. Here the gap has two beneficiaries and a price tag, which is why it was built rather than closed.
It is also worth noticing what Citi is hedging. The bank simultaneously operates its own tokenised-deposit services, participates in bank-consortium stablecoin work, and now runs settlement under an exchange's rails. Three horses, one race: whichever instrument wins the payment function, the bank intends to be its infrastructure. That is not a bet on stablecoins. It is a bet against the deposit's monopoly on moneyness.
Deposit insurance is the strongest piece of the moneyness machinery, and the sweep account shows its edge with unusual clarity. Insurance does not insure money; it insures a specific claim on a specific institution — up to €100,000 under the EU scheme, $250,000 in the United States, per depositor per bank (Directive 2014/49/EU). Coverage protects qualifying deposit claims on the insured institution; it does not follow the balance into a different instrument merely because the same dollar remains visible in the same interface.
In the swept account, the migration is measured in seconds. The incoming dollar sits inside the insured perimeter while it rests on the bank's book and outside it once it has become a custodied stablecoin — whose own backing includes deposits that are insured, if at all, for the issuer's benefit rather than the end customer's. Custodial structures can, under conditions of titling and record-keeping, pass insurance through to end customers on fiat balances; there is no analogue for the token leg. The sweep therefore changes not only the balance's economic character but the legal perimeter within which loss protection operates — presented as a single number to a customer who was told the account was deposit-style. Who ultimately absorbs losses across these adjacent regimes — bank resolution, custodian insolvency, issuer failure — is a subject this library will return to. For present purposes the narrow point suffices: the guarantee follows the claim, and the claim has left.
Step back and the deposit resolves into a bundle that was never marketed as one: a payment instrument, a safe store of nominal value, and a funding input for bank lending, fused in a single balance and cross-subsidising one another through the unpriced spread. Tokenisation is unbundling it. The stablecoin takes the payment function and runs it on a reserve with the economic characteristics of a narrow bank — fully reserved, statutorily constrained — but without deposit insurance or a lender of last resort. The yield migrates to whichever intermediary needs it as an acquisition cost. The funding function is the residual: it stays with the bank only as long as the balances do.
This is the same structural movement traced in The Shortening of Money from the duration side: instruments that decompose the deposit's bundle also decompose the behavioural assumptions banks have been allowed to make about it. An account that sweeps automatically is an account whose stability characteristics are set by code rather than by inertia — a repricing of assumptions that regulation currently hands to banks free of charge, and a subject with its own essay's worth of consequences. The deposit that remains after the unbundling is not the deposit the prudential framework was calibrated on. It is the part nobody has yet found a reason to pay you to move.
The editorial rule of this library is to start with the thing everyone thinks they understand, strip away the label, identify the legal and economic reality underneath, and ask who benefits from the gap. The deposit is the rule's limiting case, because the label is the most successful in finance. The legal reality has been public since 1848; the machinery built on top of it was explicitly designed so that the reality would never need to be consulted; and the beneficiaries of the gap included, for a long time, essentially everyone — depositors got moneyness, banks got franchise, the state got a banking system that funds itself.
What changed on 28 September is small and precise. An instrument now exists that performs the examination commercially: it takes the unexamined claim, converts it into a chain of examinable ones, and pays 3.75% a year for the privilege. The deposit is not money — not in the sense the vocabulary sells. It is a private claim that a great deal of machinery makes behave like money. What the sweep exposes is how much architecture that behaviour requires, and what happens when a competing instrument begins to price the decision to leave it. Someone is now paying you to notice — and the price of noticing has, for the first time, a public quote.
Written in a personal capacity. Analytical views only — not legal or investment advice.
Julian Gretzinger — Investor and writer on monetary history, real wealth mechanics, and financial markets. substack.com/@juliangretzinger