Instant isn't final. Payment institutions just got the thing that is.

There are two questions you can ask about a payment, and the industry has spent three years conflating them. The first is when the money moves. The second is when the movement stops being reversible by someone who is not a party to it — a liquidator, a court, an insolvency rule written long before anyone thought about ledgers.

CPMI-IOSCO keep them apart by definition. Technical settlement "describes the point in time at which the state of an SA's electronic ledger reflects that a transaction has occurred". Final settlement is something else entirely: "Final settlement (or settlement finality) is a legally defined moment."

A stablecoin transfer answers the first question very well. It does not answer the second at all. And while that has been true throughout, something changed on the other side of the comparison that has had almost no coverage.

What does settlement finality actually protect you from?

Not fraud, not error, not counterparty credit. It protects a completed payment from being unwound retroactively because somebody in the chain went insolvent.

The Settlement Finality Directive does four specific things for a participant in a designated system. Transfer orders and netting are "legally enforceable and binding on third parties even in the event of insolvency proceedings against a participant", provided they entered the system before proceedings opened — with a same-day good-faith window for orders entered afterwards, where the system operator can prove it neither knew nor should have known. Irrevocability attaches at a moment the system's own rules define: "A transfer order may not be revoked by a participant in a system, nor by a third party, from the moment defined by the rules of that system." Insolvency gets no retroactive effect, which is the provision that disapplies zero-hour rules — under which some insolvency laws deem a bankruptcy to take effect from midnight and void a whole day of payments. And collateral is unaffected by insolvency and may be realised.

A fifth protection is narrower than it is usually described. On insolvency, the participant's rights and obligations arising from its participation are determined by the law governing the system, which removes a conflict-of-laws problem — but only that one. Securities collateral still follows the law of the place of the register.

Put commercially: it is the difference between unwind risk you have to collateralise, pre-fund or insure against, and unwind risk that has been legislated away. It does not remove settlement risk. It removes the possibility of a completed settlement being reversed by a third party afterwards.

What changed for payment institutions in April 2025

Until recently that protection was, in practice, for banks. The definition of "institution" in the Directive covered credit institutions, investment firms, public authorities and publicly guaranteed undertakings, and equivalent non-EU undertakings. The legislator's own assessment was that payment institutions and e-money institutions were "effectively prevented from participating".

Article 4 of the instant payments regulation changed the definition, and article 5 of the same regulation set the date: "Member States shall adopt, publish and apply, by 9 April 2025, the laws, regulations and administrative provisions necessary to comply with Articles 3 and 4."

Three limits matter, because coverage tends to state the change without them. It applies only to systems whose business is executing payment transfer orders — not securities settlement systems, and the recital says the change is "only for the purpose of defining participants of a payment system". Firms benefiting from the small-institution exemptions, including account information service providers, are excluded. And admission is conditional on a new provision in the payment services directive requiring a description of the measures safeguarding users' funds, a description of governance and internal control arrangements including ICT, and "a winding-up plan in case of failure" — though the same article lets each member state decide how compliance is assessed, and expressly permits that to "take the form of self-assessment".

A quieter change in the same instrument may matter more. The duty requiring access rules to be objective, non-discriminatory and proportionate previously carried an exemption for designated systems. That exemption has been deleted. Designated payment systems used to sit outside that rule; since April 2025 they sit inside it.

The euro-area central bank leg opened on 6 October 2025, postponed once from June because, on the ECB's own record, of delays in national transposition. The terms repay reading. Admission is discretionary — the guideline says a central bank "may, at its discretion, also admit" non-bank providers, and the recital concedes admission "on a similar, although not equivalent, basis to credit institutions", granted "on a discretionary basis". There is no access to central bank credit. Central banks may not offer these firms safeguarding accounts. End-of-day balances are capped at twice peak daily outgoing value, with penalties on the excess.

Two things stop that being a grievance. The discretion to refuse is itself delimited: the recitals say central banks "may only refuse a request" where the applicant fails to meet the applicable requirements. And the discretionary category is not a non-bank category — investment firms, which have been institutions under the Directive since 1998, sit in the same bucket. The real line is between credit institutions and everyone else.

For comparison, the Bank of England opened direct access in July 2017, and the first non-bank joined a UK payment system settling in central bank money in April 2018. Seven and a half years ahead of the Eurosystem, and the terms are still not equivalent.

Does a stablecoin transfer settle?

Not in this sense, and nothing in European law says otherwise.

We searched the full text of the markets in crypto-assets regulation, some 637,000 characters. The word "finality" appears zero times. So do "irrevocable", "unconditional" and "definitive". The regulation covers the issuer, the reserve and the service provider; the obligations attaching to transferring crypto-assets on a client's behalf concern the client agreement and say nothing about finality. So do ESMA's level-three guidelines on the same service. The one appearance of "final settlement" is a duty on trading platform operators to initiate settlement within 24 hours on-chain, or by the close of day off-chain — a timeliness rule whose own recital concedes that on-chain settlement "might take longer as it is not controlled by the crypto-asset service provider".

The register tells the same story. ESMA's list of designated systems, in its April 2026 version, carries 135 entries across the 30 EEA jurisdictions it covers, several of them umbrella descriptions covering more than one system. We scanned every populated cell for crypto, stablecoin, token, DLT, blockchain and distributed. Zero matches. Every designated system in the EEA is a conventional real-time gross settlement, clearing or securities settlement arrangement.

This is not an oversight nobody noticed. CPMI-IOSCO addressed it in 2022, warning that stablecoin arrangements "may feature 'probabilistic settlement', where a misalignment between the state of the ledger and legal finality may occur", and that the problem "may be exacerbated in the absence of a legal entity responsible for the SA's transfer function" — because without one, there may be nobody against whom to enforce legal finality when it conflicts with what the ledger says. Their guidance asks such arrangements to define the moment of irrevocability, establish a clear legal basis for it, and prevent misalignment. It applies to arrangements authorities have determined to be systemically important. No stablecoin arrangement has been so determined anywhere we could find, so it currently binds nobody.

Protocol finality is an engineering property

The chains do use the word, and they are not being dishonest about it. They are describing something real and different.

Ethereum finalises at epoch boundaries: slots of twelve seconds, epochs of thirty-two slots, and finality requiring a supermajority link between two successive checkpoints. That is about thirteen minutes at best and, on ethereum.org's own figure, around fifteen in practice, because a transaction landing mid-epoch waits longer. The guarantee is explicitly conditional and explicitly economic. A finalised block cannot be reverted "unless there has been a critical consensus failure and an attacker has destroyed at least 1/3 of the total staked ether", which makes reversal not impossible but "economically inviable". The protocol also contemplates the chain failing to finalise for more than four epochs, and answers it with an inactivity leak that drains the stake of validators not attesting to the majority chain until finality resumes — a mechanism whose stated purpose is ensuring liveness failures are only temporary.

Avalanche's documentation carries both registers on one page. Near the top, Snowman consensus provides "sub-second, immutable finality". Further down, under a heading called Finality, "Snowman Consensus is probabilistically safe up to a safety threshold" whose failure probability "can be made arbitrarily low by adjusting system parameters". The docs treat the second as explaining the first rather than contradicting it, and they are entitled to. On an Optimism-stack chain there are three separate heads — unsafe, safe and finalized — where the finalized head is defined by reference to Ethereum's finality, and moving value back out crosses a fault-proof window on top. On Stellar, an issuer that set the clawback flag before a holder established their trustline can burn the asset out of that holder's account after the ledger has recorded the transfer.

None of this is a criticism. These are precise engineering statements. But "cannot be reverted without destroying a third of the staked ether" is a different kind of claim from "a liquidator cannot unwind this", and only the second is what a treasurer means by final.

The one DLT system that has it, and how

There is a real counterexample and it proves the point rather than undermining it. The Sterling Fnality Payment System holds settlement finality designation in the UK — on its operator's account, "the only DLT-based payment system to receive this designation" there. What it settles in is not a stablecoin: participants use "a representation of funds held at central bank".

The Eurosystem's own DLT work lands in the same place. Its bridge to TARGET, Pontes, is targeted to go live on 21 September 2026, and obtains finality by referring the cash leg back into T2: "Final settlement in central bank money for the cash leg is achieved once the corresponding transaction is completed in T2, ensuring legal certainty and robustness." When the ECB set out what it wants to add afterwards, "settlement finality on the Eurosystem DLT" was on the list of future enhancements.

Finality, so far, is not produced by a ledger. It is produced by a legal designation, and every DLT system that has it gets it by pointing back at a central bank.

Where this argument is weakest

The protection may not have been yours anyway. The Directive protects transfer orders entered by participants into a designated system. It runs to the institutions inside the system, not to their customers. A company paying a supplier by ordinary credit transfer gets no settlement finality protection either. If you are comparing a stablecoin payment with a bank transfer as a corporate, the honest comparison is much narrower than the one this article has been making, and the gap only opens once you are the institution rather than its customer.

Private law may substitute for part of it. The strongest counter-argument we found is not from the industry but from UNIDROIT, whose principles on digital assets provide that "no rights based on a proprietary claim relating to a digital asset can be successfully asserted against an innocent acquirer of that digital asset", including where control was changed by someone acting wrongfully. That addresses the competing-claimant problem without any designation or register. It does not address the liquidator: the same principles expressly preserve insolvency rules on "the avoidance of a transaction as a preference or a transfer in fraud of creditors", which is the exact risk the Directive removes. And they are soft law that works through whatever good-faith acquisition rules a state already has.

The scale argument has weakened this month. We had intended to write that the volumes make this premature. Visa disclosed on 8 September 2026 that its stablecoin settlement run rate had passed twenty billion dollars annualised, more than fifteen times higher than a year earlier, across nine chains and more than 160 stablecoin-linked card programmes — and it no longer calls the programme a pilot. Against Visa's total volumes that remains a rounding error. A rounding error growing fifteenfold in a year is not an argument for waiting. Coinbase's policy chief made the broader case against the central bankers in June: par breaks routinely in the incumbent system without anyone concluding deposits are not money — "Pay a $2 fee to withdraw $20 at an out-of-network ATM and you have taken a 10% haircut on your 'par' bank balance."

What is actually changing

The gap is being legislated at, and hardly anybody is covering it. In December 2025 the Commission proposed a settlement finality regulation that would repeal the 1998 directive and amend the financial collateral directive. The abbreviation DLT appears 47 times in it. Its impact assessment describes the preferred option as changing the definitions of system, participant and institution "to make them DLT compatible and thus to allow certain DLT-based systems to benefit from settlement finality protections", and the proposed definitions of transfer order, book-entry and account are rewritten to accommodate cryptographic signatures and decentralised records.

Read that alongside the agreed payment services text, where "funds" means central bank money issued for retail use, scriptural money and electronic money, "including electronic money tokens", and a route appears: a euro e-money token payment system could be designated and obtain settlement finality protection. Neither instrument is law; both procedures were still ongoing when we checked. The ECB has separately cautioned on a related proposal to let securities depositories settle cash legs in e-money tokens, pointing to issuer credit risk, price fluctuation and liquidity risk in the reserve assets — it does not oppose it outright, but would confine it to currencies where tokenised central bank money is not practically available.

The central bankers have been direct about the principle. Christine Lagarde in May: because stablecoins "can break away from their peg during times of stress, they do not confer the unconditional finality that central bank money does", and "a settlement layer built on private stablecoins risks weakening that principle". Isabel Schnabel at Jackson Hole in August, under a section heading that is itself the argument — stablecoins cannot be a substitute for central bank money in settlement — concluding that they "are best understood as complements to central bank money, not substitutes for it".

What we could not verify

What Qivalis, the European bank consortium's euro stablecoin venture, will run on. Crypto press reported on 8 September that it is deploying on public Ethereum. There is no primary source for it. Tracing the reports back, they rest on a single post by a third-party account, not on any Qivalis statement; Qivalis has issued no press release since 20 May 2026, and the word Ethereum appears nowhere on its site. The 37-bank figure being reported as news is real but dates from that May release. The company's own standing disclosure is the clearest statement of where it actually is: "Qivalis is not yet authorised and does not currently issue electronic money or provide payment services to the public."

Whether any named non-bank has actually joined TARGET or TIPS. The door opened in October 2025 and the ECB publishes no participant list by firm type. Named e-money and payment institutions do appear on the participant list of at least one designated system — the Lithuanian central bank lists several among CENTROlink's participants — but appearing on an operator's list is not the same as being a participant in the Directive's sense, and they were on it before April 2025, so it is not evidence of what the change did.

One question has been put, and is worth following. In February 2026 the Financial Markets Law Committee asked the Bank of England to clarify whether it expects systemic stablecoin issuers to be designated under the UK settlement finality regulations, noting that the application of those rules to distributed ledgers, "and public permissionless ledgers in particular, is currently subject to legal uncertainty". We could find no answer, and no equivalent question put to an EU competent authority about the directive.

If you are choosing a settlement layer this year, the useful question is not how fast it confirms. It is what happens to a payment that has already gone through if an intermediary in the chain fails the next morning — and whether the answer is written in a system's rulebook or in nobody's.

If you have put that question to a supervisor and received an answer, we would like to know what it was. It is the one thing that would improve this article and we could not find it.