Are stablecoins credible for payments? The BIS says no — and merchants are sitting a different exam

What the BIS actually said on August 28, 2026

Speaking at the Federal Reserve's Jackson Hole Economic Policy Symposium on August 28, 2026, Pablo Hernández de Cos — General Manager of the Bank for International Settlements, the central bank for central banks — delivered the sharpest official-sector line on stablecoins of the year: they do not provide a credible foundation for payments at scale, and should not replace conventional money in everyday transactions.

He was not calling for a ban, and it is worth being precise about that. Reporting on the remarks notes he allowed that the two instruments could coexist, with stablecoins serving more specialised roles while tokenised deposits handle the bulk of day-to-day payments. He also conceded stablecoins could become relevant if redeemability, cross-chain interoperability and integrity controls improve materially.

The substance behind the soundbite is not new, and it is not thin. It is the argument the BIS set out at length in its Annual Economic Report 2026, Chapter III — "Anchoring trust in money: innovation beyond stablecoins", and which de Cos had already framed in a speech in Tokyo in April 2026. The report's own summary is blunt: current stablecoin designs show "deviations from par value in secondary markets and limits to elasticity and interoperability," cannot "ensure exchange at par across issuers and blockchains under all conditions," and in consequence "resemble exchange-traded fund (ETF) shares rather than a means of payment."

That last line is the one worth sitting with, because it is both the most damning thing in the report and — for a business trying to get paid — the most clarifying.

The four tests, and what each one is actually measuring

Strip the speech down and there are four distinct claims, which get flattened together in headlines but pull in very different directions once you are the one holding an unpaid invoice.

Notice what happens as you read down that middle column. The tests get further and further from anything that happens at a checkout. Singleness and interoperability are about the properties of the instrument in your hands. Elasticity and bank funding are about the architecture of the financial system as a whole — genuinely important, and genuinely not a variable a hair salon or a freight forwarder controls.

Where the BIS is right, and merchants should say so

The temptation for anyone in this industry is to treat the speech as an attack and reach for a rebuttal. That is the wrong instinct. Three of the four points survive contact with reality at the merchant level, and each one has a concrete operational consequence.

Singleness: this is an argument against holding, not against accepting

The BIS's point is precise. There is no mechanism that forces one issuer's dollar token to exchange at par for another's; if you want to move between them you sell one in a secondary market and buy the other, and the price you get is the price you get. "Redeemable at par with central bank money with finality" is a standard stablecoins do not meet by construction.

The merchant translation is not "therefore don't accept them." It is: do not accumulate a shelf of different dollar tokens and call it treasury. Decide which one or two you settle in, convert on arrival, and price in the currency you actually keep books in. The "ETF shares" line is a useful discipline here — an ETF share is a fine thing to receive and a poor thing to sit on indefinitely without deciding it is what you meant to hold. This is the same discipline behind the accounting question of whether a stablecoin balance counts as cash, and behind knowing the difference between the tokens you accept: see USDT vs USDC for payments.

Interoperability: the single biggest way a merchant loses money on this rail

Here the BIS is describing a systemic fragmentation problem, but the merchant version is more mundane and more expensive. There are multiple dollar tokens, deployed across multiple chains, with the same ticker meaning different things depending on the network. A customer who sends the right token on the wrong chain has not made a payment — they have made a support ticket, and sometimes a permanent loss.

This is precisely the work a processor exists to absorb: presenting only the token-and-chain combinations you actually accept, generating a fresh address or QR per sale so nothing is ambiguous, matching the exact amount received against the exact invoice, and flagging underpayment and overpayment rather than silently swallowing the difference. Fragmentation is real. It is also a solved problem at the checkout layer, and pretending it solves itself is how merchants get hurt.

Financial integrity: the rail changes, your obligations don't

De Cos's concern is that AML controls are applied inconsistently across stablecoin platforms and jurisdictions. At the level of global policy that is a fair criticism and an active work programme. At the level of your business it produces exactly one instruction, and it is the same instruction we give in every vertical we write about: customer identification, invoicing, record-keeping, sanctions duties and tax are unchanged. A different settlement instrument does not narrow the file you are required to keep.

What it does change is the quality of the evidence. An on-chain settlement is a timestamped, independently verifiable record of an exact amount arriving at an address you control, matched to an invoice that names what was sold — plus a full audit log on the processor side. Compared with an envelope of cash or a wire whose intermediaries you cannot see, that is a compliance asset, not a compliance gap. Which is a narrower claim than "crypto is compliant," and it is the only one worth making.

Where the frame stops reaching the merchant

The fourth test is the one that does not transfer, and the reason is structural rather than rhetorical.

Elasticity is about a system's ability to create liquidity on demand — for the BIS, the thing that "underpins the functioning" of money's foundational properties. A commercial bank creates a deposit when it makes a loan; a stablecoin issuer cannot. That is a real and important difference if you are designing a monetary system. It is not a question a business asks of its acceptance method. When you invoice a client, you are not asking the payment rail to extend credit into existence. You are asking it to move an agreed amount from them to you.

There is one honest exception, and it deserves to be stated plainly rather than buried: if your customers buy from you specifically because a card gives them credit, a stablecoin payment does not replace that. A card is two products fused together — a transfer instrument and a consumer lending product. Stablecoins replace the first and not the second. For a business whose average ticket depends on instalments or revolving credit, that is a reason to keep cards, not to drop them. Every merchant we work with runs this rail alongside their existing ones, for the customers the existing ones turn away.

The same asymmetry applies to the "no-questions-asked" standard the BIS invokes — money that passes hand to hand indefinitely without anyone assessing the issuer. That is a demanding test for a bearer instrument in general circulation. A merchant's exposure is one identified customer, one invoice, and a holding period measured in minutes before conversion. That is not zero risk, and we have written about the counterparty risk you do carry when you hold an issuer's token. But an instrument you hold for four minutes is not being asked to be money. It is being asked to be a transfer.

The number that reconciles both sides: 0.2%

The strongest evidence for the BIS position is not rhetorical, it is arithmetic. In a July 8, 2026 address, the Financial Stability Board's Deputy Secretary General Martin Moloney put stablecoin cross-border payment volume at less than 0.2% of total cross-border payments in 2025, against roughly US$200 trillion in total cross-border flows. His conclusion was measured rather than dismissive: their "most immediate value may be as components in hybrid models — integrated with bank money and interoperable FX/settlement — rather than as stand-alone global rails," while cautioning that even that "is [not] more than a possibility at this point." You can read the full remarks in the FSB's "Cross-Border Payments: Towards the Next Chapter".

Take that number seriously. At 0.2%, stablecoins are not the payment system. The BIS is right that nothing about current volumes justifies redesigning monetary policy around them, and anyone telling a merchant that cards are about to be replaced is selling something.

Now look at the same number from the other end. 0.2% of US$200 trillion is on the order of US$400 billion a year, and it is not spread evenly across the world's commerce like a thin film. It concentrates in precisely the corridors and customer types where the incumbent rail refuses, delays or deducts: payers without a local bank account, countries under capital controls or FX rationing, weekend and holiday settlement, correspondent chains that take five days and arrive short.

A merchant does not sell into "global payment flows." A merchant sells to a queue of specific people, and the only question that matters is whether the ones in front of them can pay. A rail carrying 0.2% of the world's money can still carry 100% of the customers your bank turns away. Both statements are true, and they are not in tension — they are just measured at different altitudes.

Tokenised deposits solve the BIS's problem. They don't solve yours.

The constructive half of the speech is the recommendation: bring tokenisation inside the two-tier system. Tokenised deposits are account-based bank liabilities settled between banks across central bank accounts, which preserves singleness by design. As monetary architecture, this is coherent, and it is very likely where the official sector lands.

But look at what that design is, from the counter. A tokenised deposit is a claim on a bank, transferred between account holders at participating banks. Improving how banks settle with each other is a genuine improvement — for people who have bank accounts at participating banks. The customer who fails your current rail overwhelmingly fails because that is the exact thing they do not have: the newly relocated member with no local account, the importer in an FX-rationed corridor, the relative in another country trying to contribute to a funeral, the traveller whose foreign debit card cannot carry a deposit hold. Making bank-to-bank settlement faster does nothing for a payer who is outside the perimeter entirely. That is not a flaw in the design; it is what the design is for.

Then there is the calendar. We covered the two flagship bank-led efforts this year: the largest US banks' shared tokenised-deposit network through The Clearing House, planned for mid-2027, and the BankChain Alliance announced by 39 state bankers associations on August 25, 2026 — targeting 2027, with no technology partner selected and no individual bank publicly committed as an owner. Both are serious. Neither can accept a payment for you this quarter.

So the honest framing of the choice is not "stablecoins versus tokenised deposits." It is: there is a customer who cannot pay you today, and the officially preferred fix is dated 2027 and structurally addressed to somebody else.

One more piece of the BIS argument deserves acknowledgement rather than a dodge, because it is the one with real teeth in emerging markets: dollar-token adoption can accelerate digital dollarisation, weakening monetary sovereignty and raising bank funding costs. That is a serious policy concern and we treated it at length when the IMF's own First Deputy Managing Director made the same warning in August. It is also not a merchant's lever. A shop in Buenos Aires quoting in dollars is not causing dollarisation — it is responding to it, years after its customers did.

How Payzum fits the honest version of this

Everything above narrows to a short product spec, and it happens to be the one Payzum was built to. Read it as a set of constraints derived from the BIS's own criticisms, not as a rebuttal of them.

• Non-custodial, so there is no issuer-plus-processor stack of risk. Funds settle directly to wallets the merchant controls. Payzum never holds, pools or controls the money — the settlement is the payment. That removes the failure mode the BIS worries about least and merchants worry about most: an intermediary balance that can be frozen or lost. It is the same argument as a non-custodial processor generally.
• Issuer-agnostic, so singleness is a choice you make once. Accept crypto broadly, then use optional auto-conversion to USDC or USDT so what lands is the dollar token you decided to hold — not whichever one the customer happened to have. You are not running a cross-issuer FX book by accident.
• Fragmentation handled at the checkout, not by the customer. Bitcoin, Ethereum, Solana, Polygon, Base, Arbitrum, Optimism, BNB Chain and Avalanche are supported, with typical confirmations around 0.4s on Solana and ~2s on Base and Polygon. A fresh QR per sale, invoices with expiry and overpayment detection, and signed webhooks mean the chain-and-token question is answered before the payer sees it.
• Finality, stated with both edges. On-chain settlement is final: no chargebacks, no reversal window, no acquirer holding a reserve. That is the headline benefit and also an obligation — there is no scheme to arbitrate a dispute, so your refund policy has to be written down and honoured, and refunds are payments you initiate.
• A record, not a hole. 2FA, encrypted secrets, signed webhooks and a full audit log, alongside KYC in the product. The compliance file you keep is the same file; the settlement evidence in it is better.

What Payzum does not do, and will not claim: it does not settle to fiat bank accounts, it is not a bank, it is not a custodian, and it does not change a single regulatory obligation you already carry.

How it works, step by step

1. Pick the flows that are actually broken. Not "accept crypto" in the abstract — the specific customer segment your current rail refuses. Foreign cards at the counter, cross-border invoices, weekend settlement, recurring billing to customers who move country, payouts to contractors abroad.
2. Connect a wallet you control and set the settlement policy. Choose which chains and tokens you accept and whether incoming payments auto-convert to USDC or USDT. This is the singleness decision, made once, in a dashboard, instead of accidentally every time a customer pays.
3. Turn on the instruments that match the flow. Payment links and buttons with no code; hosted checkout as redirect, modal or inline; invoices with expiry and overpayment detection; recurring subscriptions; a tip jar or donations page; and POS with a fresh QR per sale, PIN-protected cashiers and per-cashier analytics, where any phone is a terminal.
4. Wire it into what you already run. REST API with API keys, signed webhooks and an integration playground, plus drop-in compatibility with existing e-commerce plugins. If you sell an API to AI agents, x402 lets agents pay in USDC on Base per call, straight to your wallet — you configure your existing endpoint and price, no protocol to implement.
5. Pay out on the same rail. Mass payouts by CSV and EVM stablecoin payouts across Polygon, Arbitrum, Optimism, Base, BNB Chain and Avalanche, so the money going out doesn't fall back to the rail you just routed around. See crypto mass payouts.

Where this shows up in real businesses

The abstract argument gets concrete fast. In each of these, the BIS's macro critique is entirely correct and entirely beside the point, because nobody involved is trying to replace the monetary system — they are trying to close one transaction.

• The customer with no local bank account. A coworking operator signs a member who relocated three weeks ago. They have no local account and no local card, so the resident-only instant-payment scheme is closed to them and the cross-border card autopay will silently die on the next reissue. A tokenised deposit network, whenever it arrives, will require exactly the account they don't have. A subscription billed to a wallet renews regardless. This is the same shape as getting paid from abroad without a bank account.
• The invoice that arrives short. An exporter bills an importer on another continent against a sailing date. The wire clears in five days minus correspondent deductions, and the customer's ledger says paid in full while yours says short. An invoice with expiry, a reference tied to the container, and overpayment detection turns a monthly argument into a reconciliation line.
• The counter where every card is foreign. A restaurant, gallery booth or charter base whose customers arrived by plane pays cross-border interchange plus FX on nearly every sale, and takes a dispute months later from another country. A QR per sale settles in seconds to the owner's wallet with no reversal window — the argument behind accepting crypto in person.
• The API that sells to machines. An AI agent cannot open a merchant account, will not remember a password, and needs to buy one call for a fraction of a cent. There is no tokenised-deposit design on any roadmap that addresses a buyer with no legal personality. That market is being served today, in USDC on Base.

Waiting for tokenised deposits vs accepting stablecoins today

Objections worth taking seriously

"If the BIS says this isn't credible money, why would I touch it?"

Because you are not being asked to treat it as money. Businesses routinely accept instruments that fail every one of the BIS's tests — gift cards, store credit, vouchers, loyalty points, trade credit — and nobody demands singleness from a gift card, because nobody is proposing it as the anchor of a monetary system. The relevant question for an acceptance method is narrower: does the payment arrive, is it final, does it land somewhere you control, and can you convert it to what you keep books in. On those four, this rail answers cleanly. On "should this be the base money of an economy," the BIS's answer is no, and we have no argument with it.

"Isn't the par-value risk real, though?"

Yes, and it should be sized honestly rather than dismissed. You are exposed to an issuer's ability to hold its peg for as long as you hold its token. The mitigations are boring and effective: settle in tokens you have actually diligenced, convert on arrival rather than accumulating, avoid holding a shelf of issuers you never chose, and treat the balance as a receivable you have decided to keep rather than as cash by default. That is a real risk with a real management practice — not a reason to leave the money uncollected, and not something to wave away either. We go through it properly in counterparty risk in stablecoin payments.

"Won't regulators just close this once tokenised deposits arrive?"

Nothing in the BIS position calls for that, and the regulatory direction in 2026 has run the other way — toward defining which dollar tokens are permissible and under what conditions, rather than removing them. De Cos explicitly described coexistence, with stablecoins in specialised roles. What a merchant should take from it is a design principle rather than a forecast: stay issuer-agnostic and non-custodial, so that a newly favoured regulated token is something you add in a settings screen, not a migration.