Custody & Infrastructure11 min read
MB
Editorial Team
·November 17, 2026

How Did Project Meridian FX APAC Synchronise Three Ledgers?

Project Meridian FX APAC is a 2026 experiment in which the Bank of England, the Monetary Authority of Singapore and the Bank of Thailand used one Synchronisation Operator to settle both legs of FX trades across three different test systems: a simulated RTGS, a centrally managed token ledger and a wholesale CBDC shared ledger. The reports, published on 13 November 2026, show that gross and multilateral net payment-versus-payment worked, and that technical atomicity did not settle the question of legal finality.

TL;DR — Key Takeaways

  • ✓The test: On 13 November 2026 the Bank of England, MAS and the Bank of Thailand published Project Meridian FX APAC: one Synchronisation Operator, built by Partior, settled FX legs across a simulated RTGS, a centrally managed token ledger and a wholesale CBDC shared ledger.
  • ✓The mechanism: The operator matched both legs, reserved funds directly on two systems, used a dedicated escrow account on the third, and instructed settlement only once every leg was secured, so the legs settled together or not at all.
  • ✓The results: Real-time gross PvP and multilateral net PvP both worked. In every exception scenario tested, the operator either completed settlement or fully unwound it. Non-overlapping operating hours and earmark expiry were not tested.
  • ✓The gap: The policy report says technical atomicity does not by itself establish a common, legally recognised moment of finality, and that rules on who may pause, cancel or retry a linked transaction still need to be written.
  • ✓The UK link: The Bank of England says the findings will inform its live synchronisation service, targeted for 2028. MAS and the Bank of Thailand make no equivalent commitment.

Ready to get started?

Join others who are already using our platform.

How Did Project Meridian FX APAC Synchronise Three Ledgers?

What did Project Meridian FX APAC test in 2026?

It tested whether one Synchronisation Operator could settle FX payment-versus-payment (PvP) across three central bank test systems built on different designs: an account-based RTGS, a centrally managed token ledger and a shared ledger for wholesale CBDC.

The Bank of England, the Monetary Authority of Singapore (MAS) and the Bank of Thailand (BoT) ran the project. The Reserve Bank of Australia and Bangko Sentral ng Pilipinas took part as observers in the policy workstream. Partior, a DLT-based market infrastructure provider for clearing and settlement, was the delivery partner. According to the MAS media release of 13 November 2026, it developed, integrated and tested the Synchronisation Operator (SO) prototype. The findings sit in three Bank of England documents: an executive summary, a technical report and a policy report.

The problem it addresses is quantified in the MAS release. Citing the 2025 BIS Triennial Central Bank Survey, it says more than US$1.4 trillion, about 10% of average daily FX settlement in April 2025, remained fully exposed to settlement risk. PvP removes that exposure by making one currency move only if the other does.

Table 1 of the technical report sets out the three systems. None of them shares a ledger with another.

SystemDesignHow the SO connectedHow funds were reserved
Bank of England Synchronisation LabSimulated account-based RTGS with a prototype earmarking capabilityREST over HTTPS, JSON; SO polls for statusEarmark on the participant's account
Bank of Thailand Project SANPrototype token-based system with a centrally managed token ledger and an interoperability layerISO 20022 XML over HTTPS via the interoperability layer; status by callbackEscrow within the SAN arrangement, released conditionally, without the SO taking custody
MAS SGD Testnet (non-production)Shared ledger with participant-controlled wallets for wholesale SGD CBDCJSON-RPC to the ledger; SO listens for ledger eventsTokens moved into a technical escrow wallet held by the SO, under the bank's authorisation

This is the difference from the earlier Project Meridian FX report of 24 April 2025, in which the BIS Innovation Hub and Eurosystem central banks linked an emulated UK RTGS to three Eurosystem interoperability solutions. It is also different from the European route described in the analysis of how Pontes bridges DLT platforms into TARGET, where one central bank system is extended to reach the ledgers. Meridian FX APAC leaves each system as it is and puts a co-ordinator between them.

How did one Synchronisation Operator settle both legs of an FX trade?

It ran a four-stage process: settlement instruction, matching, funds reservation and settlement. Before the first settlement instruction executes, the trade can be cancelled by releasing the reservations; after it, the operator only moves forward.

Each participant sent the SO an instruction for the sell side of its FX trade: debtor, creditor, amount, currency, value date and a priority flag. The SO matched the two legs by computing a SHA-256 digest over the debtor agent, amount, creditor agent, currency and settlement date and time. A matched pair at normal priority went into a queue for netting; any other matched pair started a bilateral PvP workflow at once.

The technical report splits the bilateral flow into two phases around a “point of commitment”. In the reversible phase, the SO asks each system in turn to earmark funds. Earmarked funds cannot be used by the account holder unless the SO cancels the earmark. If any earmark fails, the SO releases those already obtained and stops. Once every leg is earmarked, the SO has reached the point of commitment. In the irreversible phase it issues settlement instructions and, if a system goes down, keeps retrying rather than reversing.

“The operator reserved funds directly on two systems, while operating a dedicated escrow account to hold funds temporarily on the third. […] Once all legs were secured, the operator instructed atomic settlement, ensuring that the payments settled together or not at all.”

Bank of England, Project Meridian FX APAC: Technical Report, section 2

The prototype was a centralised system, not a DLT application. It had five parts: an API layer for banks, a UI layer for monitoring, a matching engine, a workflow orchestrator that enforced atomic execution, and an adapter layer that absorbed the differences between the three central bank interfaces. The core settlement logic was written once; each adapter handled polling, callbacks or ledger events for its own system.

Why did the operator need an escrow account on SGD Testnet?

Because the SGD Testnet configuration had no third-party earmark, the SO replicated reservation by moving tokens into a dedicated bank-owned escrow wallet it managed under the bank's authorisation. Both reports call this experiment-specific, not part of a future SO model.

The three systems differed in what they let a third party do. The Synchronisation Lab and Project SAN recognised the SO as an instructing party at the interface. SGD Testnet required a cryptographic signature from the account holder for every asset movement. To avoid holding participant keys, the adapter built each unsigned instruction, routed it to the participant's external key custodian for signature and then submitted it. The SO held only its own operator key.

Project SAN also used escrow as its reservation mechanism, but within the SAN arrangement itself. The technical report says this replicated a technical account “without the SO taking custody of the funds”. The policy report names only the escrow in MAS' test environment as an additional function that goes beyond co-ordination.

“In the case of SGD Testnet, the SO assumed responsibilities that would not typically arise in a conventional synchronisation model where SOs co-ordinate settlement without holding funds.”

Bank of England, Project Meridian FX APAC: Technical Report, section 6

The policy report explains why the distinction matters. An SO that temporarily holds or controls funds raises custodial, credit, liquidity, safeguarding and insolvency questions that do not arise when funds stay in participants' accounts and are only earmarked. For readers who have built single-platform settlement, this is the same trade-off covered in the architecture of cross-chain DvP with permissioned token escrow, where one party locks both legs. Meridian FX APAC shows the cost of falling back to escrow when the underlying ledger cannot reserve funds on its own.

How did multilateral net PvP work across three ledgers?

The SO queued eligible trades, calculated each participant's net position per cycle, and negotiated earmarks system by system until all three systems accepted the same batch. In the prototype, a failed earmark usually led to removing that participant's smallest transaction and retrying.

A cycle ran on a schedule or on demand. It locked a snapshot of queued trades into a batch, with both legs of every trade in the same batch. Each participant then needed a single debit or credit per system, however many counterparties it had. Because the SO could not know in advance which trades every system would accept, earmarking became an iterative negotiation across the three systems.

Failure information differed by system, which changed how fast the negotiation went:

  • Synchronisation Lab: reports status per movement record, so the SO could see which debits were waiting for liquidity and attribute them to specific participants.
  • Project SAN: reports failure per instruction; where the SO could attribute the failure to a participant, it removed that participant's smallest transaction.
  • SGD Testnet: reports failure per participant instruction, and the netted positions ran as a sequence of individual movements.

In all three systems the SO calculated the net positions itself. The technical report says that an SO calculating offsetting positions “may begin to assume functions traditionally associated with a clearing house”. It suggests one alternative: central bank systems that accept multilateral net settlement messages, such as pacs.029, and perform the netting themselves. The reports give no figure for how much liquidity the netting saved, and the experiment used configurable balances and a limited set of simulated transactions rather than realistic volumes.

What happened when a leg failed?

In all four exception scenarios tested, the SO either completed settlement or fully unwound it, without leaving a single leg to settle alone. Before the point of commitment it rolled back; after it, it retried.

ScenarioPhaseSO response
1. Insufficient fundsReversibleReleases the first leg's earmark, terminates, records the reason on both legs; never retried
2. Invalid account identifierReversibleCancels any earmark on the other leg and terminates
3. Net cycle with both errorsReversibleIsolates both failing trades, cancels their earmarks, reruns the cycle with the rest
4. Central bank system outageIrreversibleNo rollback; retries at increasing intervals until the system returns or the operator-intervention threshold is reached

The technical annex lists six failure types treated as definitive and never retried, including LIQUIDITY_FAILURE and INVALID_REQUEST_PAYLOAD. Every other failure, such as a timeout or an unparseable response, is treated as transient. If an outage outlasts the maximum retry horizon, the SO holds the workflow in a persistent state with earmarks intact and alerts a human operator. Deterministic instruction identifiers made retries idempotent, so a receiving system would recognise a repeat as a duplicate.

Two conditions that matter most across Asian and European time zones were outside scope. The technical report says the experiment “did not examine non-overlapping operating hours for central bank systems or earmark expiry”. It assumed overlapping system availability. One idea it lists for future testing is to keep an earmark open overnight in one system while another reopens.

Does atomic settlement across the three systems equal legal finality?

No. The policy report states that “technical atomicity does not by itself establish a common, legally recognised moment of finality across participating systems.” Each domestic system keeps its own final settlement event, and no common finality model was chosen.

The policy report uses Principle 8 of the Principles for Financial Market Infrastructures (PFMI), which requires an infrastructure's rules to define the point at which settlement becomes final, and Continuous Linked Settlement (CLS) as reference points. CLS records PvP settlement on one common ledger. An SO instead co-ordinates settlement inside separate domestic systems under their own rules and laws. A live arrangement would have to define when a transaction becomes irrevocable in each system and how those moments combine into one cross-border outcome.

“The technical workflow's point of commitment and retry logic can support operational integrity but cannot resolve that legal question by themselves.”

Bank of England, Project Meridian FX APAC: Policy Report, section 4c

This is the same split between operational and legal finality set out in the explainer on when a tokenized transfer is actually final, now applied across three jurisdictions at once. The governance question is still open as well. The policy report says rulebooks, contracts and operating procedures must allocate “who may approve, pause, cancel or retry transactions”, how persistent failures are handled and how losses are allocated. None of that exists yet for a cross-border SO.

How would regulators classify a Synchronisation Operator?

By what it does, not what it is called. The policy report takes a non-custodial SO with four co-ordination functions as its baseline, and says holding funds or calculating net positions moves it towards custodial and clearing-type functions with more questions attached.

The four baseline functions are instruction processing, funding co-ordination, readiness monitoring, and settlement and exception co-ordination. Even without holding funds, the SO creates a co-ordination dependency. Table 1 of the policy report says a failure at the SO “could delay or prevent linked settlement even where the underlying central bank settlement systems remain available.”

The report organises the open questions into five lenses: access and assurance, regulatory perimeter and oversight, settlement finality, governance and accountability, and data and information sharing. It cites existing models as partial reference points, not ready answers:

  • Access: the RBA's Batch Administrator framework for RITS, which separates access to settlement infrastructure from access to central bank money and supports reservation batches.
  • Oversight: co-operative central bank oversight of Swift, led by the National Bank of Belgium, with the caveat that an SO plays a more active role in execution than a messaging provider.
  • Finality: PFMI Principle 8 and CLS, as described above.
  • Governance: PFMI Principle 2 and the Global Layer One Market Infrastructure Toolkit.
  • Data: ISO 20022 messaging, which standardises format but not what data an SO may receive or the lawful basis for sharing it.

The report's split is practical: access, assurance and perimeter questions sit mainly with each jurisdiction, while finality, accountability and co-operation between authorities arise from the cross-border design. It suggests internationally developed principles for SOs “if considered useful” and recommends no single service design.

What does Meridian FX APAC mean for tokenised asset settlement?

It shows that cash on one central bank system and value on another can be settled together if each system lets a third party reserve and release funds. For tokenised assets the main use is the UK: the Bank of England has committed to a live synchronisation service targeted for 2028.

The commitment was announced in the FCA and Bank of England release of 18 May 2026 and repeated in a footnote to the Bank of England's Meridian FX APAC release. The executive summary adds that participation “does not imply an equivalent implementation commitment or a particular policy direction” by MAS or the Bank of Thailand. Delivery-versus-payment is listed only as possible further work; the reports contain no DvP result.

For a token issuer or venue whose asset lives on its own ledger, the experiment turns into four design requirements on the asset side, mirroring Table 2 of the technical report:

  • Authorised third-party access: the ledger must let a named operator, distinct from the holder, act on a holding.
  • Earmarking: a holding can be locked against a pending settlement without moving it to someone else's wallet, which avoids the custody questions the SGD Testnet escrow raised.
  • Separate settlement instruction: release of the earmarked holding is a distinct, unambiguous step, so the trade stays reversible until every leg is ready.
  • Status and idempotency: queryable transaction status, reliable event delivery and duplicate-safe instruction IDs, which the technical report links to avoiding outcomes where a leg's status cannot be established.

Permissioned token standards with freeze and partial-freeze functions can supply some of this, but a freeze imposed for compliance and an earmark for settlement are different holds with different release rules. A token layer that records the reason for each hold, and who may release or unwind it, maps onto the governance gap the policy report describes. The custody consequences of who controls those holds are covered in the guide to institutional RWA custody and compliance requirements.

Who this matters for: settlement and treasury teams planning tokenised asset or deposit settlement in the UK against the 2028 target, and infrastructure teams in Singapore or Thailand watching the SGD Testnet and Project SAN. Who it does not yet help: anyone needing a production service today, cross-time-zone settlement, or a legal finality opinion. The reports were built in test environments, and the policy report frames all of its analysis without presuming “a particular implementation pathway or commitment to live deployment.”

Frequently Asked Questions

What is a Synchronisation Operator?

A Synchronisation Operator (SO) is a co-ordination layer that makes linked transfers in separate settlement systems happen together or not at all. In Project Meridian FX APAC it matched the two legs of each FX trade, asked each central bank system to reserve the funds, and only then instructed settlement. In the baseline model of the policy report, the SO holds no participant funds and keeps no settlement accounts.

Which central banks took part in Project Meridian FX APAC?

The Bank of England, the Monetary Authority of Singapore and the Bank of Thailand ran the experiment. The Reserve Bank of Australia and Bangko Sentral ng Pilipinas joined as observers and contributed to the policy workstream. Partior, a DLT-based clearing and settlement provider, was the delivery partner that developed, integrated and tested the SO prototype. The findings were published on 13 November 2026.

Is Project Meridian FX APAC a live service?

No. It ran in simulated or non-production test environments with stylised participants, configurable balances and a limited set of simulated transactions. The policy report says any live arrangement would need assessment without presuming a particular implementation pathway or commitment to live deployment. MAS and the Bank of Thailand state no implementation commitment. Only the Bank of England has a separate plan: a live synchronisation service targeted for 2028.

Does atomic settlement in Meridian FX APAC mean legal settlement finality?

No. The policy report states that technical atomicity does not by itself establish a common, legally recognised moment of finality across the participating systems. Each domestic system defines its own final settlement event. A live arrangement would have to explain how those events combine into one definitive outcome for the linked trade, including during outages, and no finality model was chosen in the reports.

Did Meridian FX APAC test delivery-versus-payment for tokenised securities?

No. The experiment tested only the two central bank money legs of FX trades: real-time gross PvP and multilateral net PvP. The executive summary lists delivery-versus-payment as possible further technical work. The Bank of England says the findings will inform its development towards a live synchronisation service, but the reports contain no DvP results.

Ready to get started?

Join others who are already using our platform.