🌐 Cross-border payment rails around the world

Why no payment really crosses a border

A so-called cross-border payment is a chain of domestic payments made in at least two countries, not a single transfer that crosses a border. The reason is that there is no global central bank, and so no settlement asset shared by the peso and the won. The domestic legs are tied together by entries on the books of institutions that extend credit to each other. Somewhere along the chain, the currency is converted. Every link runs its own compliance checks. The four problems the literature keeps coming back to (cost, speed, transparency, and access) all stem from this one design. The “rails” covered here are the different ways of stitching those domestic payments together.

• Correspondent banking, the traditional chain of banks, which covers every currency and amount and is sped up by Swift gpi.
• Multilateral regional platforms such as PAPSS, Buna, AFAQ, SADC-RTGS, REPSS, EAPS, and SIPA, which pool the correspondent function across a region.
• Bilateral links between instant payment systems such as PromptPay–PayNow, UPI–PayNow, and PayNow–DuitNow: two domestic real-time rails connected to each other.
• Multilateral interconnection through Nexus: one connection instead of n × n bilateral agreements.
• Private networks: money transfer operators (Western Union, Ria, Remitly), fintechs (Wise, Nium, Payoneer, XTransfer), push-to-card (Visa Direct, Mastercard Move), wallet gateways (Alipay+), and mobile money hubs (Onafriq).
• Tokenized settlement, with Partior, Fnality, Citi Token Services, Circle Payments Network, Ripple Payments, and mBridge’s cross-border CBDCs.

Swift and correspondent banking, the universal layer

S.W.I.F.T. SC is a member-owned cooperative incorporated under Belgian law. It was founded in 1973 and went live in 1977. It connects more than 11,500 institutions in more than 200 countries and territories and carried more than 13.4 billion FIN messages in 2025, or about 53.3 million a day (Swift, 2025 annual review). No other infrastructure comes close to that reach, so even rails built to bypass Swift end up relying on it. CIPS, China’s system, still uses Swift for messaging on more than 80% of its flows.

The mechanism underneath is correspondent banking. The sender’s bank holds an account with a bank in the destination country. It calls this its nostro account, meaning “our account with you.” The bank that holds the account calls the same account a vostro account. The payment is a series of debits and credits across these accounts. No money actually travels. Each link takes a cut: a processing fee, sometimes a lifting fee that an intermediary deducts from the amount in transit, and a markup built into the exchange rate. That FX markup is almost always the largest cost, and it is the only one the customer never sees.

This network is shrinking. The number of active correspondent relationships fell by about 30% between 2011 and 2022, even as volumes grew (BIS, Bulletin No. 87, 2024, based on 2023 CPMI data). De-risking means dropping a correspondent relationship that no longer pays for the compliance risk it brings. Small jurisdictions in the Pacific, the Caribbean, and East Africa are hit first. Some regions built their own rails for exactly this reason, not out of political ambition. Pacific Transfers, run by Digicel Pacific between Fiji, Samoa, Tonga, and Vanuatu, exists because correspondent banks pulled out of the region.

gpi, Swift Go, and the retail framework: speed, and where it gets lost

gpi (global payments innovation) is a service rulebook layered on top of existing correspondent banking, not new infrastructure. Participating banks commit to same-day crediting, fee transparency, and passing remittance data along in full. The UETR is what makes these commitments verifiable. It is a unique reference carried in the message header since November 18, 2017. It lets the payment be traced from the sender to the final beneficiary without the identifier getting lost on the way, and it feeds the Tracker that banks along the chain check.

Two offerings extend gpi to consumers. Swift Go, launched in 2021, targets low-value payments by consumers and small businesses, with fees and delivery times disclosed up front. On March 5, 2026, Swift and a group of banks announced a framework for cross-border retail payments. It rests on four commitments: fees and the exchange rate disclosed up front, the full amount delivered to the beneficiary, handling of the last mile, and end-to-end visibility. The announcement lists corridors to Australia, Bangladesh, Canada, China, Germany, India, Pakistan, Spain, Thailand, the UK, and the US. The plan calls for more than 25 banks live by the end of June 2026 and more than 70 institutions signed up over the year (Swift, March 2026). Swift openly presents the framework as a defensive move against money transfer fintechs.

• Require a UETR on every outgoing instruction and store it in your records. Without it, no inquiry can be investigated and customer service spends its days on the phone.
• Require structured party addresses in the pacs.008. Free-text addresses are the leading cause of false positives in sanctions screening, and therefore of delays.
• Pick the charge code deliberately: OUR (the sender pays all fees and the beneficiary receives the full amount), SHA (shared, so intermediaries take their fees along the way), or BEN (the beneficiary pays). A biller that expects an exact amount cannot work with SHA.
• Get your bank to pass on the Tracker statuses, not just an “executed” flag. They are the only proof of the date the beneficiary was credited.
• Check the cut-off time of the correspondent and of the receiving bank, including local public holidays. An instruction sent after the Dubai cut-off on a Thursday evening lands on Sunday or Monday, depending on the country.
• Record the exchange rate applied and its spread over the mid-market rate. It is the largest and least visible cost in the chain.

FX settlement: CLS, PvP, and orphan currencies

Herstatt risk is the risk that, in an FX trade, one counterparty delivers the currency it sold but never receives the currency it bought. It is named after the German bank whose failure in 1974 left counterparties that had delivered Deutsche marks without the dollars they were owed. Every cross-border payment that involves a conversion includes an FX trade, and the time difference between the two settlement systems involved is enough to open that window of exposure. The mechanism that closes it is PvP (payment versus payment). Both legs settle at the same time, or neither does.

CLSSettlement, run by CLS Bank International since 2002, is the global tool for this. It settles more than $8 trillion in payments a day across 18 currencies. More than 75 settlement member institutions and more than 38,000 third-party users take part (CLS Group, product page accessed in July 2026). The operator’s legal status has a consequence that rarely gets noticed. CLS Bank is a New York Edge Act corporation, supervised by the Federal Reserve and designated as systemically important by the FSOC. PvP settlement of the Canadian dollar or the Mexican peso therefore falls under US regulatory jurisdiction, whatever the currency involved.

A newer generation of infrastructure aims to eliminate the gap between the two legs rather than hedge it. Partior, a joint venture of J.P. Morgan, DBS, and Temasek set up in Singapore in 2021, offers a shared ledger for 24/7 atomic settlement of cross-border payments and interbank FX PvP. Settlement is simultaneous and final, with no intraday exposure. Fnality runs the Sterling Fnality Payment System, live since 2023, with an omnibus account at the Bank of England. It backs a distributed ledger with central bank money. The central bank issues no token in this setup and only holds the collateral. Citi Token Services, live since October 2024, applies the same principle within a single banking group. None of the three discloses volumes.

Regional platforms: sharing the correspondent

A regional cross-border payment platform is infrastructure shared by several countries in the same region. It clears and settles payments between their banks. Regions turn to one when correspondent banks pull out, or when routing through the dollar makes trade within the region more expensive. The design replaces N bilateral relationships with one shared platform, settles in local currencies, and converts only the net balance into hard currency. Western payment teams know little about these rails, even though they are often the only workable way to collect funds in southern Africa, the Levant, or Central America.

PAPSS is the most closely watched regional platform of the group, and the most politically charged. Afreximbank designed and funded it and remains its settlement agent. Through PAPSS, a Nigerian bank can pay a Kenyan bank in naira and shillings without going through a dollar correspondent outside Africa. When the BEAC joined on July 9, 2026, it brought in all six CEMAC countries at once, with their banks expected to go live by the end of 2026. A link with Kenya’s Pesalink was announced in February 2026. On July 7, 2025, the group unveiled PACM (PAPSS African Currency Marketplace), an order-book FX marketplace for African currency pairs, built with Interstellar on a permissioned distributed ledger. PAPSS estimates that routing through hard currencies costs $5 billion a year, and that African airlines alone have more than $2 billion in blocked funds.

In the Americas, Directo a México shows the gap between a rail existing and a rail being used. This official ACH corridor from the US to Mexico, run by the Federal Reserve Banks as part of FedGlobal since 2003, uses a Banco de México reference rate and charges very low fees. Yet it never caught on against money transfer operators, because it lacked distribution and consumer awareness. On November 25, 2025, Federal Reserve Financial Services announced it was shutting down FedGlobal ACH Payments. Payments to Mexico will no longer be accepted after November 20, 2026. A cheap, technically superior public rail can disappear simply because nobody uses it.

Bilateral instant payment links: the Asian model

A bilateral instant payment link connects two national instant payment systems directly, allowing account-to-account transfers between the two countries. Southeast Asia’s central banks built the first of these links by plugging their domestic rails into each other. The flow is simple. The payer enters a mobile number or scans a QR code. The system resolves the alias on the other country’s rail, and an FX quote appears before the payer confirms. Each leg then runs as a domestic payment. There is no correspondent, no lifting fee, and no uncertainty about the amount received.

Nexus, mBridge, and CIPS: what is coming, and what is geopolitics

Nexus is a scheme for multilateral interconnection of instant payment systems, built to solve the n × n problem. It was developed by the BIS Innovation Hub, which completed the blueprint in July 2024, and then handed over to Nexus Global Payments (NGP). This Singapore nonprofit was incorporated on March 26, 2025, by the central banks of India, Indonesia, Malaysia, the Philippines, Singapore, and Thailand. Each instant payment system connects just once, to the hub, and can then reach every other member, with a target of cross-border payments in under 60 seconds. The FSB puts the target go-live date in 2027 (consolidated report, October 2025).

Europe has the legal framework for this kind of connection, but it has seen no significant use so far. The European Payments Council’s OCT Inst (One-Leg Out Instant Credit Transfer) scheme has been in force since 2023 and runs on EBA CLEARING’s RT1. It covers only the European leg of an instant payment whose other leg is outside SEPA or in another currency. Actual adoption is still low. IXB (Immediate Cross-Border Payments) was the effort by EBA CLEARING, The Clearing House, and Swift to connect RT1 with the US RTP network on the EUR/USD corridor. It never got past the pilot stage: a proof of concept in October 2021, then a pilot launched in October 2022. No go-live has been announced since, and the service is not in the current catalog. In a product plan, IXB counts as a pilot, not an available service.

Project mBridge is the world’s most advanced cross-border CBDC (central bank digital currency) project, and it is not a Western one. Its consortium brings together the HKMA, the PBoC’s digital currency research institute, the Bank of Thailand, the Central Bank of the UAE, and the Saudi Central Bank. The project reached minimum viable product in 2024. The BIS Innovation Hub then pulled out, and the participating central banks carried on alone. mBridge settles peer to peer in digital central bank money, with no correspondent and no third-party messaging. In other words, it bypasses both chokepoints: the dollar and Swift.

Private rails: MTOs, fintechs, cards, hubs, and tokenization

Private rails are cross-border transfer services run by commercial companies, alongside the public clearing and settlement infrastructures. They carry a large share of global flows, though that share is poorly measured. Their shared economic principle is to keep the money from crossing the border. The operator holds accounts or liquidity on both sides, collects locally, pays out locally, and settles only the net balances. Wise and Bitso both follow this model, differing only in the settlement asset.

Real costs and delivery times, the G20 roadmap, and what breaks in production

The G20 roadmap for cross-border payments is the work program launched in October 2020 to make these payments faster, cheaper, more transparent, and more inclusive. Quantitative targets for the end of 2027 were added in October 2021, with the remittance cost target aligned with Sustainable Development Goal 10.c for 2030. Five years on, the FSB sees progress on speed for wholesale payments. Cost, however, has not budged. The FSB considers it unlikely that satisfactory improvements will be achieved globally on the planned timeline (consolidated report, October 9, 2025).

• The cut-off time, not the rail. An instruction sent after the correspondent’s or the receiving bank’s cut-off waits until the next business day. Friday evenings in the Levant and the Gulf are a classic case, since the weekend there does not line up with Europe’s.
• Sanctions screening on unstructured addresses. The leading cause of multi-day delays in correspondent banking, and the reason ISO 20022 introduced structured addresses.
• Currencies outside CLS. PvP covers only 18 currencies. For everything else, someone carries settlement risk for several hours, and who that is belongs in your corridor data.
• The limit on a bilateral link. Southeast Asia’s instant links cap amounts per day and per beneficiary, which rules out B2B use cases.
• The regional rail’s actual bank coverage. A rail present in 28 countries is not present at every bank in those 28 countries. Only the named list of live participants tells you the real coverage.
• The hidden FX markup. On most corridors, it is larger than the explicit fees. To measure it, compare the rate against the mid-market rate at the time of the transaction, not against the rate the provider displays.
• The nonbank last mile. Mobile wallets, cash pickup, and agents are where the failure rate and the real delivery time are decided. No international rail contract covers this part.
• Mistaking an announcement for a launch. Nexus, PACM, and BRICS Pay are three announced projects, and none is open for live use. A product plan that assumes they are available will reach its target date with no rail to carry its flows.