Fintech
Central Banks Test Atomic FX Settlement Across Three Simulated Systems

Key PointsBespoke adapters enabled cross-system interoperability but risked expanding the…
View more
- Bespoke adapters enabled cross-system interoperability but risked expanding the operator beyond a coordination role.
- Technical atomicity across separate ledgers does not establish legal settlement finality across jurisdictions.
- The non-live prototype left operating hours and prolonged outages unresolved for any production service.
The Bank of England, Monetary Authority of Singapore (MAS) and Bank of Thailand announced on 13 November 2026 that a Project Meridian FX APAC prototype had coordinated simulated bilateral and multilateral net foreign-exchange (FX) settlement on a payment-versus-payment (PvP) basis across three materially different central-bank wholesale test environments. The work addresses a market in which more than US$1.4 trillion, or 10%, of average daily FX settlement in April 2025 was settled on a gross bilateral basis and remained fully exposed to settlement risk.
Gross bilateral settlement leaves counterparties exposed to loss of the full trade value. PvP removes that risk by making the final payment of one currency conditional on the final payment of the other.
Meridian FX APAC tested whether a Synchronisation Operator (SO) could preserve that all-or-nothing outcome across separate infrastructures rather than requiring a common ledger. The prototype matched currency legs, arranged funds reservation and instructed settlement only after every required leg was ready. The results indicate that central banks can modernise their systems on different technologies without making interoperability contingent on a single infrastructure design.
Three Architectures Required Different Adapters
Partior developed, integrated and tested the SO prototype. The technical report identifies the Bank of England’s contribution as its Synchronisation Lab with a simulated real-time gross settlement (RTGS) environment, the Bank of Thailand’s as its Project SAN tokenised payment prototype with an interoperability layer, and MAS’s as a non-live configuration of SGD Testnet, a shared ledger for wholesale central bank digital currency. Adapters accommodated differences in third-party access, funds reservation, settlement instructions and status reporting.
Those differences changed what the operator had to do. It reserved funds directly in two systems but used a dedicated bank-owned escrow wallet under participant authorisation in the MAS configuration to replicate reservation. That was an experiment-specific adaptation, not an inherent feature of synchronisation. It also captures the central trade-off: bespoke integration enabled interoperability but added complexity and could move a live operator beyond a lightweight coordination role.
Netting and Failure Tests Define the Trade-Offs
For multilateral net settlement, non-urgent matched transactions were queued into cycles and the operator calculated each participant’s net position. A common batch then had to be accepted and reserved across all three systems before settlement. If one system could not reserve sufficient liquidity, the operator could release existing reservations, remove an affected transaction and retry with a revised batch, potentially reducing liquidity needs while preserving atomic settlement.
Exception tests covered insufficient funds, invalid account identifiers and temporary infrastructure outages. Across the selected scenarios, the operator either completed settlement or fully unwound the transaction without allowing an individual payment leg to proceed independently. After the technical point of commitment, an outage triggered repeated settlement attempts rather than a rollback, because reversing one leg while another settled would undermine the PvP mechanism.
Technical Atomicity Is Not Legal Finality
The work remained a non-live proof of concept using stylised participants, configurable balances and a limited set of simulated transactions. It did not reproduce the full FX trade lifecycle, live participants or production operating conditions. The tests also left non-overlapping operating hours, time-zone differences, earmark expiry and prolonged outages for further work, material qualifications for a service intended to coordinate liquidity across jurisdictions.
The policy report distinguishes technical atomicity from legal settlement finality: coordinating payment legs does not itself establish a common, legally recognised settlement outcome across separate domestic systems and jurisdictions. It assesses the operator through five lenses: access and assurance; regulatory perimeter and oversight; settlement finality; governance and accountability; and data and information sharing. Existing frameworks provide building blocks, but not a ready-made model for synchronisation.
The baseline policy model is non-custodial, meaning the operator coordinates settlement without holding participants’ funds. Temporarily controlling funds or calculating net positions could nevertheless raise custody, safeguarding, insolvency and clearing-related questions. The adaptations that made the test work can therefore change the operator’s legal characterisation and risk profile, making system design a regulatory question as well as an engineering one.
The findings will inform the Bank of England’s work towards a live RTGS synchronisation service. The Bank of England and Financial Conduct Authority’s joint Call for Input on UK wholesale markets targets delivery in 2028, with further design, regulatory, assurance and application information planned for early 2027.
The executive summary states that participation does not imply an equivalent implementation commitment or particular policy direction by MAS or the Bank of Thailand. Further experiments could add jurisdictions and industry participants, more currencies and transaction scenarios, delivery-versus-payment use cases, and a wider range of operating and liquidity conditions.
The project consequently separates technical feasibility from live-service readiness. Different infrastructures can be coordinated without converging on one design, but a production service still depends on legally robust finality, resilient operating arrangements and liquidity controls that work beyond a controlled experiment.











