Market cap $2.20 T −3.64 % over 24 hours24h volume $71.83 BBTC dominance 59.4 %ETH 10.1 %Fear & Greed 10 Extreme fearCoins 22,053

Three central banks settle simulated FX trades across tokenised and legacy ledgers, and only the Bank of England sets a 2028 date

Bank of England, Singapore and Thailand settled simulated FX trades across three ledgers in Meridian FX APAC: what the test proved and what is still open.

12 min read

Abstract globe with stacks of coins, a bank building and a bridge between them
Table of contents
  1. Synchronised settlement: how it removes FX settlement risk
  2. What three central banks actually tested
  3. How a trade settled: four stages and a point of no return
  4. Settle or unwind: the four failure tests
  5. Multilateral netting worked, but the operator ended up doing a clearing house’s job
  6. What is still open: legal finality, time zones, outages and 2028
  7. From Meridian to Meridian FX APAC: the earlier steps
  8. Why it matters for stablecoins and tokenised deposits
  9. What to watch next

The Bank of England, the Monetary Authority of Singapore (MAS) and the Bank of Thailand published the results of Project Meridian FX APAC on Friday 13 November 2026: a test in which one coordinating operator settled simulated foreign exchange trades across three central bank money systems built on different technologies, one account-based, one tokenised and one a wholesale CBDC ledger. According to the Bank of England’s executive summary, the two currency legs of each trade settled together or not at all, including when the researchers injected failures. Only the Bank of England tied the work to a live date: it says it is committed to launching synchronisation in 2028.

The test ran in simulated environments, with stylised banks instead of real ones, a limited set of transactions and no production conditions. Neither of the two published reports gives a settlement time, a throughput figure or a measure of the liquidity that netting saved. What they do give is a detailed account of how the plumbing worked, where it needed workarounds and which legal questions are still unanswered.

Partior, the company that built and tested the operator prototype, was the technical partner. The Reserve Bank of Australia and the Bangko Sentral ng Pilipinas joined the policy workstream as observers.

Synchronised settlement: how it removes FX settlement risk

A currency trade has two legs. A bank in London pays pounds, a bank in Singapore pays Singapore dollars. If one leg settles and the other does not, one side has paid and received nothing. That is settlement risk, often called Herstatt risk after the German bank closed in June 1974 after its counterparties had paid Deutsche Marks but before the dollars it owed in New York went out. Payment-versus-payment (PvP) removes the risk by making both legs settle at once. Synchronisation is the Bank of England’s way of doing PvP without moving everyone onto one shared ledger: an outside operator, the Synchronisation Operator (SO), locks the funds in each domestic system, checks that every leg is ready, then tells each system to settle.

The scale of the problem is measurable. The BIS Quarterly Review of June 2026, drawing on the 2025 Triennial Survey, found that more than $14 trillion of gross obligations were settled on an average day in April 2025. Just over $5 trillion (36%) went through PvP. More than $1.4 trillion (10%) was settled gross and bilaterally, which the BIS describes as fully exposed to settlement risk.

FX settlement, average day, April 2025 (BIS)AmountShare
All gross obligations settledmore than $14 trillion100%
Settled via payment-versus-paymentjust over $5 trillion36%
Gross bilateral, fully exposed to settlement riskmore than $1.4 trillion10%

What three central banks actually tested

The technical report lists three systems with very different designs, all used in simulated or non-production form:

  • The Synchronisation Lab of the Bank of England, a test environment that imitates a classic centralised, account-based real-time gross settlement (RTGS) system, with a prototype function to earmark cash for an outside operator.
  • Project SAN, from the Bank of Thailand, a prototype tokenised payment system with a central token ledger and an interoperability layer that routes instructions to the settlement layer.
  • The MAS SGD Testnet, a shared-ledger infrastructure for issuing, transferring and redeeming wholesale Singapore dollar CBDC, where banks control their own wallets and there is no central intermediary. It is an operational environment, but the project used it only in a non-live configuration.

The operator itself was a centralised system, not a blockchain. It had an interface for banks to send instructions, a matching engine, a workflow engine that enforced the all-or-nothing rule and an adapter layer that translated its requests into whatever each system understood. The core settlement logic was written once; the differences lived in the adapters.

Those differences were real. The Bank of England lab speaks REST and JSON and has to be polled until a transaction changes state. Project SAN takes ISO 20022 messages and calls the operator back when a step is done. The Singapore ledger is reached through JSON-RPC and announces completion as a ledger event that the operator has to listen for. The Singapore ledger also wants the account holder’s cryptographic signature on every movement, so the operator built unsigned instructions, sent them to each bank’s own key custodian for signing, and held only its own operator key.

Abstract server rack linked by chains and plugs to a bank building and gear
One operator, three central bank systems: an account-based RTGS, a tokenised ledger and a wholesale CBDC testnet, each reached through its own adapter.

How a trade settled: four stages and a point of no return

Every trade went through the same sequence. Each bank sent the operator its side of the deal. The operator matched the two legs by computing a SHA-256 fingerprint of the paying bank, the amount, the receiving bank, the currency and the settlement date and time, and looking for the mirror instruction. It then asked each system to earmark the funds, which freezes them so the account holder cannot spend them elsewhere. Once every system confirmed, it sent the settlement orders.

The design hinges on what the report calls the point of commitment. Before it, any failure means the operator releases every earmark and the trade dies with no money moved. After the first settlement order goes out, the operator never tries to reverse anything. If a system goes dark at that moment, it keeps retrying, at growing intervals, until the system comes back or a human operator is called in.

Reserving funds did not work the same way everywhere. On the Bank of England lab the operator placed earmarks directly in participants’ accounts. On Project SAN funds sat in escrow inside the SAN arrangement and were released on instruction, which the report says replicated a technical account without the operator taking custody. On the Singapore testnet, which has no earmarking for third parties, tokens moved into a dedicated bank-owned wallet that the operator managed under the bank’s authorisation, acting as an escrow. The executive summary is clear that this last workaround is not a proposed model for a future operator.

Settle or unwind: the four failure tests

The researchers fed the prototype four kinds of trouble. When one bank lacked funds, the operator released the other leg’s earmark and cancelled the trade, and both banks went back to where they started. When an instruction carried an account number that did not exist, the trade was cancelled at once and never retried. When one netting cycle contained both errors at the same time, the operator isolated each failing transaction and ran the cycle again without them. When a system dropped offline after the point of commitment, the operator kept retrying rather than unwinding, because an unwind at that stage could leave one leg settled and the other not.

The report sorts failures into two families. Six error types, including a liquidity failure and an invalid payload, are final and never retried. Everything else, such as a timeout, a lost connection or an unreadable reply, counts as temporary. “The SO was able either to complete settlement successfully or to fully unwind the transaction,” the technical report says. It credits three properties of the connected systems: idempotency (a repeated instruction is processed once), a status that can be queried and reliable delivery of events.

Multilateral netting worked, but the operator ended up doing a clearing house’s job

The second use case batched non-urgent trades into cycles and settled each participant’s net position across all three systems, so that a bank pays or receives one amount instead of many. This is how liquidity gets saved. The hard part, the reports say, was not the arithmetic but getting three systems to agree on the same set of trades while the money stayed locked.

The prototype solved it by negotiation. It asked each system in turn to earmark the batch; when one refused, it used whatever failure information that system returned to identify the bank at fault, usually removed that bank’s smallest transaction, and tried again until all three accepted. How fast that works depends on the detail a system reports: the Bank of England lab flags individual movements, while Project SAN and the Singapore ledger report failure at the level of a whole instruction.

The catch is that none of the three systems did the netting itself. The operator calculated the net positions and sent the results. In the technical report’s words, an operator in that position “may begin to assume functions traditionally associated with a clearing house”. One option it suggests is for central bank systems to accept multilateral net messages and net the trades themselves, which would keep the operator light. Savings from netting were not quantified.

The policy report is where the limits show. Its baseline is a non-custodial operator that coordinates but never holds money. The escrow wallet in Singapore and the netting calculations both pushed the prototype beyond that line, and the report says such functions can raise custodial, credit, liquidity, safeguarding and insolvency questions. “Policy treatment should depend on the functions performed by the SO in practice rather than how the SO is classified or described,” it says.

The biggest gap is finality. Each domestic system defines its own moment at which a payment becomes irrevocable. The operator’s point of commitment is a technical event, not a legal one, and the policy report says technical atomicity “does not by itself establish a common, legally recognised moment of finality”. It compares the setup with CLS, the existing cross-border FX settlement system: CLS records settlement on one common ledger, whereas a synchronisation operator would coordinate settlement taking place in separate national systems under separate laws.

Several real-world conditions were left out on purpose. The test assumed every system was open at the same time, so it did not model London, Singapore and Bangkok working different hours, nor what happens when an earmark expires overnight. Long outages were not fully tested. Liquidity was simplified to configurable balances, and no trade was settled in commercial bank money. One idea the technical report floats for future tests is holding an overnight earmark in one system until another reopens, compared against common settlement windows.

Abstract scales of justice beside an hourglass, a calendar and a warning sign
Legal finality, different operating hours and earmark expiry were left open by the test, while the Bank of England aims for 2028.

The policy report sorts the work into five lenses: access and assurance, regulatory perimeter and oversight, settlement finality, governance and accountability, and data sharing. It borrows from existing models, among them the Reserve Bank of Australia’s Batch Administrator framework for access, the cooperative oversight of Swift led by the National Bank of Belgium, and PFMI Principle 8 on finality. Its own conclusion is that none of them covers the whole problem.

On commitments, the Bank of England says it will use the findings to build a live synchronisation service, “in line with its commitment to launch synchronisation in 2028”. The same page says participation “does not imply an equivalent implementation commitment” by the MAS or the Bank of Thailand. Next technical tests named are delivery-versus-payment, wider operating and liquidity conditions, modelling of market participants and more jurisdictions.

From Meridian to Meridian FX APAC: the earlier steps

The project is the third in a line. The first Project Meridian, run by the BIS Innovation Hub London Centre with the Bank of England, introduced the idea of an operator linking the RTGS to other infrastructures and reported in April 2023. Project Meridian FX, published on 24 April 2025, added the Banque de France, the Deutsche Bundesbank, the Banca d’Italia and the ECB and linked an emulated UK RTGS with three Eurosystem settlement solutions. None of the euro area central banks took part in the APAC round.

The APAC round was announced on 13 November 2025, so the results came exactly a year later. “To realise the potential of tokenised financial systems, international cooperation is needed,” Kenneth Gay, the MAS chief fintech officer, said at the launch. The Bank of Thailand is also one of the central banks behind mBridge, the wholesale CBDC platform the BIS handed over to its participating central banks in October 2024.

A different BIS project, Agorá, published a prototype report on 27 May 2026 that puts tokenised commercial bank deposits and tokenised central bank reserves on one shared platform. The Meridian FX APAC policy report names Agorá and Nexus as reference points for its governance questions. The two approaches answer the same question in opposite ways: Agorá brings the money onto one ledger, synchronisation leaves each currency at home and coordinates between them.

Why it matters for stablecoins and tokenised deposits

Neither report mentions stablecoins. Still, this is the official sector’s answer to the same problem private tokens claim to solve in cross-border payments: moving value between currencies at once and without counterparty risk. A dollar stablecoin does it by putting one private token on one public chain. Meridian keeps final settlement in central bank money and lets each country pick its own technology, from a decades-old RTGS model to a wholesale CBDC ledger.

For banks that plan to issue tokenised deposits, the finding that counts is that interoperability did not require a common design. A bank token on one ledger could in principle settle against central bank money in a conventional system, provided both expose the same few functions: third-party access, earmarking and clear status. The Bank of England prefers banks to innovate with tokenised deposits rather than private stablecoins, Journal du Coin noted on 14 November. That preference sits alongside commercial bank projects such as Project Faro, where 21 banks back a dollar stablecoin, and the euro area debate on wholesale settlement that surfaced in the ECB Appia consultation.

For crypto markets there is no direct effect. No token is involved, nothing is live and no private firm is committed to use the service. The relevance is longer term. A live link between central bank systems, which only the Bank of England has scheduled, would give banks a route to settle both legs of a currency trade in central bank money, the same cross-border job dollar stablecoins are used for today.

What to watch next

The MAS hosts the Singapore FinTech Festival from 18 to 20 November, the first public stage for officials to comment on the results. No month has been given for the 2028 launch, and the Bank of England has not said which currencies or partners the live service would start with. The open technical items are operating hours and earmark expiry; the open legal one is a settlement moment that courts in the UK, Singapore and Thailand would all recognise.

Sources

  1. Bank of England, Project Meridian FX APAC: Executive summary and introduction, 13 November 2026 bankofengland.co.uk
  2. Bank of England, Project Meridian FX APAC: Technical Report, 13 November 2026 bankofengland.co.uk
  3. Bank of England, Project Meridian FX APAC: Policy Report, 13 November 2026 bankofengland.co.uk
  4. BIS Quarterly Review, Uncovering FX settlement risk: new measures from the 2025 BIS Triennial Survey, 15 June 2026 bis.org
  5. Bank of England, BoE, MAS and BoT to explore synchronised FX settlement across borders, 13 November 2025 bankofengland.co.uk
  6. Journal du Coin, Meridian FX APAC article, 14 November 2026 journalducoin.com

Leave a Reply

Your email address will not be published. Required fields are marked *