Module 5

Cross-Border: Where Rails Collide

Correspondent banking, SWIFT gpi, the limitations of domestic RTP networks, and how international payments will work across multiple incompatible rail systems

---

The Correspondent Banking Model Is Fundamentally Broken

When you send money across an international border today, it moves through correspondent banking. The correspondent banking network is essentially a web of bilateral relationships between banks. Bank A maintains an account at Bank B (called a "nostro" account). Bank B maintains an account at Bank A (called a "vostro" account). When Bank A needs to send money to a counterparty in Bank B's country, it debits its account at Bank B and credits the counterparty's account at Bank B. In theory, this should work seamlessly. In practice, it is a model designed for batch processing and has proven catastrophically inefficient for real-time commerce.

The original correspondent banking architecture assumed that banks would settle large batches of transactions once per day during specific settlement windows. A US bank would accumulate all transactions destined for European banks, send them in a single batch at 5 pm, and the European banks would process them overnight. The model worked in the 1970s. It does not work in the 2020s when businesses expect payment confirmation within minutes or hours, not days.

Correspondent banking moves money between countries through bilateral bank relationships and batch processing windows. It was engineered for a world where international settlement happened once per day. Cross-border payments now take three to five days despite the fact that the underlying technology (the internet) moves information instantaneously.

Here is what happens when you send $1,000 across borders through correspondent banking. You initiate the payment from your bank. Your bank sends it through the SWIFT network (the messaging system for interbank transfers). The message reaches the intermediary bank. The intermediary bank routes it to the receiving country. The message reaches the receiving bank. The receiving bank credits the account. But none of this happens instantly. Each step happens during specific processing windows. Your bank might send the payment at 3 pm and it hits the SWIFT network at 4 pm. The intermediary bank processes it at 9 am the next day. The receiving bank processes it at 2 pm. The recipient sees the money at 3 pm. Three days later, the payment is "completed."

The delays are not technical. They are operational. Banks have deliberately designed batch processing windows because maintaining continuous settlement infrastructure is expensive. For a bank to offer truly real-time international payments, it would need to maintain constant connectivity with every correspondent bank, with instant netting and settlement. This is possible but has never been economically rational for banks when batching reduces operational costs by 90 percent.

---

The True Cost of Cross-Border Payments

When you send $100 across a border through correspondent banking, the costs are opaque but substantial. You see a single fee from your bank ($15 to $50 depending on the bank and corridor). But the true cost is hidden.

The receiving bank deducts an additional fee ($10 to $30). An intermediary bank (if the payment routes through one) deducts another fee ($5 to $20). The foreign exchange conversion adds 2 to 3 percent margin on top of the real market rate. The receiving bank adds another 1 to 2 percent margin. By the time $100 leaves your account, the recipient sees $70 to $85. The 15 to 30 percent loss is not a mistake. It is the built-in cost of the correspondent banking model.

For large institutional transfers (over $100,000), the costs are partially negotiated and sometimes lower on a percentage basis. But for small transfers under $10,000 (which includes most cross-border freelancer payments, family remittances, and small business invoices), the cost structure is economically absurd. A startup paying a contractor $500 loses $50 to $75 in fees and currency margin. The contractor receives $425 to $450. Both parties lose. The system is not optimised for small cross-border payments. It is barely functional.

ISA fees (International Settlement Account fees) are another hidden layer. Correspondent banks charge fees to banks that maintain accounts for international transfers, often $500 to $2,000 per month per account, regardless of whether the account is actively used. These fees are paid to maintain relationships for rare transfers. For smaller banks, correspondent banking is so expensive that they cannot afford to maintain direct relationships with banks in all countries. They are forced to use intermediaries, adding more hops and more fees.

---

SWIFT gpi: The Improvement Layer (But Still Not Fast Enough)

SWIFT released SWIFT gpi (Global Payments Innovation) in 2017 as an attempted fix to correspondent banking. SWIFT gpi added transparency to cross-border payments: you can now track where your payment is in the correspondent chain. SWIFT gpi mandated faster processing: payments that previously took three to five days now often clear in one to two days. SWIFT gpi added end-to-end traceability: the sending and receiving banks can see the full payment route.

SWIFT gpi is an improvement, but it is still fundamentally a patch on the correspondent banking model. It made the infrastructure faster by making processing windows shorter, not by fundamentally changing the architecture. You still cannot send money instantly. You still lose money to hidden fees and currency margins. You still cannot predict exactly when the money will arrive. SWIFT gpi is like installing a faster elevator in a building when what you needed was to move to a different building entirely.

Banks adopted SWIFT gpi because it required minimal infrastructure changes. It is built on top of existing correspondent banking. But adoption is not universal. Not all banks have implemented SWIFT gpi. Many payments still route through the old non-gpi infrastructure and take five to seven days. From a customer perspective, the improvement is marginal.

---

Real-Time Payment Networks Are Domestic-Only by Design

The real-time payment networks built in the past decade (FedNow in the US, Faster Payments in the UK, UPI in India, PIX in Brazil) are engineered as domestic-only systems. They were designed to move money instantly within a single country. They do not have cross-border capabilities because building cross-border capabilities requires solving the currency conversion problem, the regulatory problem, and the network interoperability problem all at once.

This creates a fundamental gap. A merchant in Brazil using PIX (real-time) cannot easily pay a supplier in the US using FedNow (real-time). The two networks do not talk to each other. There is no automated way to route a payment from PIX to FedNow. The merchant must go through a correspondent bank or a specialised cross-border payment provider, which defeats the purpose of using real-time networks.

The domestic-only design made sense for the network operators. Building a domestic real-time network required solving problems like participant bank connectivity, fraud prevention, and settlement rules. Adding cross-border complexity would have delayed launch by years and added massive regulatory coordination challenges. But the trade-off is that real-time networks have created new fragmentation. You now have dozens of real-time networks globally, none of which can talk to each other across borders.

The Interoperability Gap

When you initiate a payment from FedNow (US) to a UK bank, the payment does not move on FedNow to Faster Payments (UK). It converts to correspondent banking. The payment leaves FedNow, goes to the receiving US bank, converts to a correspondent transfer, routes through SWIFT, arrives at the UK bank, and eventually shows up in Faster Payments. You have combined the worst of both worlds: you get the instant settlement of FedNow for the US portion (a few seconds) and the three-day delay of correspondent banking for the cross-border portion.

The interoperability problem is architectural. Real-time networks use different protocols, settlement mechanisms, and participant structures. Building interoperability between two real-time networks requires both networks to agree on message formats, settlement rules, and how to handle failures. This is feasible for two networks in adjacent countries (the US and Canada could theoretically link FedNow and Canada's Payments Networks). But building a global mesh of interoperable real-time networks is a coordination problem that has never been solved at this scale.

---

The Emerging Cross-Border Initiatives: Nexus, P27, ASEAN, EPI

Several initiatives are attempting to solve the cross-border interoperability problem by linking real-time payment networks. These initiatives are not trying to replace correspondent banking globally. They are trying to link specific networks in specific corridors where the volume justifies the effort.

Nexus: The BIS-Led Initiative

Nexus, led by the Bank for International Settlements, is an ambitious project to create a platform that links real-time payment networks across central banks. Nexus would allow a payment initiated in one country's RTP network to flow directly to another country's RTP network. It is still in pilot stages with a small group of central banks, but if it succeeds, it could fundamentally change cross-border payments.

Nexus solves the interoperability problem by creating a standardised intermediary. Instead of requiring each network to integrate directly with every other network, each network connects to Nexus. Nexus handles the protocol translation, the currency conversion (if needed), and the settlement coordination. A payment from FedNow would flow through Nexus to the UK's Faster Payments. The recipient would see the money within minutes, not days.

P27: Nordic Integration

P27 is a project to create a shared real-time payment system for Nordic countries (Norway, Sweden, Denmark, and potentially others). Instead of maintaining separate domestic RTP networks, the Nordic countries are exploring a unified payment infrastructure. P27 would allow instant cross-border payments between Nordic countries at the same speed and cost as domestic payments within a single country.

ASEAN Payment Network

The ASEAN Payment Network (APN) is attempting to link the real-time payment networks of Southeast Asian countries. Each ASEAN member has its own domestic RTP network (or is building one). APN seeks to create interoperability between them. The initiative is in early stages and coordination challenges are significant (ten countries with different regulatory frameworks), but the intent is clear: remove the friction of correspondent banking for intra-ASEAN transactions.

EPI: European Integration

The Euro Payments Initiative (EPI) is a European project to create a unified payment system across the EU for domestic and intra-European transactions. EPI is attempting to reduce European dependence on card networks (particularly Visa and Mastercard) and create a European-owned payment infrastructure. While EPI is technically distinct from real-time payment network interoperability, it serves a similar purpose: enabling faster, cheaper cross-border payments within a region.

---

Stablecoins as a Cross-Border Bridge

An alternative approach to cross-border payments is using stablecoins as settlement infrastructure. A merchant in Brazil could convert reais to USDC, send the USDC to a US merchant, and the US merchant could convert USDC back to dollars. The transaction would settle within minutes on the blockchain. The cost would be substantially lower than correspondent banking (typically $1 to $5 regardless of transaction size).

Stablecoin-based cross-border payments are already happening. Circle's USDC and PayPal's PYUSD are actively used for cross-border settlement by merchants and payment providers. The advantage is speed and cost. The disadvantage is regulatory uncertainty, the requirement that both parties have stablecoin infrastructure in place, and the volatility risk (USDC is pegged to the dollar, but the peg could break).

For high-volume corridors where both countries have stablecoin infrastructure, stablecoins are already cheaper than correspondent banking. Over the next five years, expect stablecoins to capture an increasing share of cross-border volume, particularly for B2B payments and remittances in corridors where both sides of the market have adopted stablecoin infrastructure.

Ripple and XRP have positioned themselves as a cross-border payment solution for two decades. XRP as a bridge currency promises faster, cheaper cross-border settlements than correspondent banking. However, XRP adoption for actual cross-border payments remains marginal. Banks have been reluctant to hold XRP as a bridge currency due to price volatility and regulatory uncertainty. Most banks that work with Ripple use the infrastructure company's capabilities for payments, but they use stable currencies or fiat settlement, not XRP.

Stablecoins are solving cross-border payments for corridors where both parties have infrastructure. For corridors without stablecoin infrastructure, correspondent banking remains the default, and real-time networks do not help because they are domestic-only.

---

ISO 20022: The Data Standard That Enables Everything

Beneath all these initiatives is a critical enabler: ISO 20022, a new international messaging standard for financial transactions. ISO 20022 replaces SWIFT MT standards (which have been in use since the 1970s) with a richer, machine-readable format that carries more data about each transaction.

SWIFT's old MT standards could carry basic information: sender, receiver, amount, date. ISO 20022 carries structured data: invoice details, remittance information, tax IDs, regulatory compliance data, all in a machine-readable format. This richness enables end-to-end process automation. An invoice payment no longer requires manual reconciliation. The system reads the remittance information automatically and applies the payment to the correct invoice.

The migration to ISO 20022 is mandatory in many countries. The European Union, UK, and others have mandated that financial institutions migrate from MT standards to ISO 20022. This creates a window where the messaging layer is being rebuilt. While the messaging layer is being rebuilt, the opportunity exists to rebuild the entire cross-border payment infrastructure on top of the new standard.

ISO 20022 is a necessary but not sufficient condition for cross-border payment modernisation. Having a better messaging standard does not automatically solve the correspondent banking problem or network interoperability. But it enables new solutions. Payment infrastructure built on ISO 20022 can be designed to interconnect more easily. Regulatory compliance can be automated. Cross-border payments can carry enough data context that each intermediary understands the full transaction, not just the immediate hop.

---

The Routing Decision: Which Rail for Cross-Border in 2027?

By 2027, a payment provider or merchant will not face a binary choice between correspondent banking and a single alternative. They will face a routing decision: which rail should this payment use? The answer depends on several factors.

Corridor and Geography

If both sending and receiving countries have mature real-time payment networks linked through Nexus or a similar initiative, route through the RTP network. The payment will settle in minutes at low cost.

If both parties have stablecoin infrastructure in the same stablecoin, route through stablecoin. The payment will settle in seconds at minimal cost.

If neither real-time network interoperability nor stablecoins are available, fall back to correspondent banking (or SWIFT gpi for faster settlement and fee transparency).

Amount and Urgency

For high-value payments where working capital cost matters, the routing favours faster rails. A $1 million wire that saves one day of interest is worth more than the cost difference between correspondent banking and faster alternatives. The merchant will pay more to get the money faster.

For low-value payments where cost per transaction is the constraint, the routing favours cheaper rails. A $100 cross-border payment through stablecoin ($1 fee) beats correspondent banking ($15 to $30) regardless of speed.

Regulatory and Compliance Requirements

Some cross-border corridors carry regulatory requirements that favour certain rails. Sanctions screening, AML compliance, and tax reporting are mandatory in most jurisdictions. Different rails handle compliance differently. Correspondent banking has the most mature compliance infrastructure. Newer rails are building compliance in. The routing decision must factor in which rail can satisfy the compliance requirements at the lowest cost.

Counterparty Capabilities

A payment provider cannot use a rail that the counterparty does not support. If the receiving bank in Thailand does not support stablecoins, you cannot route through stablecoin even if it would be cheaper. If the receiving country does not participate in Nexus, you cannot use the linked RTP network. The routing is constrained by the installed base of capabilities on both sides of the corridor.

---

What's Needed for Cross-Border Payment Modernisation

Solving cross-border payments requires more than technology. It requires regulatory coordination, network investment, and market-level adoption. Several pieces must fall into place simultaneously.

First, central banks must link their real-time payment networks (Nexus or similar initiatives must succeed and expand). Second, regulators must create a standardised compliance framework so that payments moving between countries do not require re-compliance at each hop. Third, payment providers must build multi-rail infrastructure so that they can route optimally based on corridor, amount, and urgency. Fourth, both sending and receiving parties must have access to non-correspondent-banking infrastructure (whether RTP networks or stablecoins).

Progress is happening but remains slow. Nexus is in pilot phase with a small group of central banks. P27 is in planning. EPI is still building. Most of the world's cross-border payment volume still moves through correspondent banking. But the trajectory is clear. Over five years, an increasing share of cross-border volume will shift to faster, cheaper rails as infrastructure matures and adoption spreads.