Why Cross-Border B2B Payments Are Still Broken (And What's Actually Fixing Them)

A mid-sized manufacturer in Vietnam paying a component supplier in Germany in 2026 is likely still waiting two to three business days for funds to arrive, paying an FX spread of 1.5-3% above the interbank rate, receiving no real-time tracking of where their money is, and paying correspondent banking fees they can’t easily itemize. This is not a niche edge case. It is the modal experience for B2B cross-border payments under roughly $5 million — a volume tier that encompasses the overwhelming majority of international business transactions by count, even if not by total value.

The persistence of this dysfunction is not for lack of attention. The fintech industry has spent a decade and several hundred billion dollars in venture capital attacking the cross-border payment problem. SWIFT has launched gpi, ISO 20022 migration is underway, and a generation of specialist B2B payment platforms has emerged. Some things have genuinely improved. Much has not. Understanding which problems have been solved, which are structural, and which solutions are actually gaining traction is the prerequisite for making intelligent decisions about international payment infrastructure.

The Correspondent Banking Chain Problem

The foundational issue with cross-border B2B payments is architectural. When a business bank in Hanoi sends funds to a business bank in Frankfurt, those two banks almost certainly do not hold accounts with each other. The payment travels through a chain of intermediary banks — correspondent banks — each of which holds accounts with the next link in the chain. Each correspondent bank processes the transaction according to its own schedule, applies its own fee, and updates its own ledger in sequence.

The number of hops in this chain varies by corridor. A USD payment between well-connected major banks might traverse one or two correspondents. A payment in a less liquid currency between smaller banks in emerging markets might traverse four or five. Each hop introduces delay — not because the technical processing takes long, but because each correspondent bank has compliance checks, batch processing windows, and business-hours constraints that pause the transaction. The net effect is settlement measured in days rather than seconds.

The fee structure is equally problematic. Each correspondent bank charges a transit fee, typically $10-30 per transaction, which is deducted from the payment principal as it passes through. A payment of $10,000 might arrive as $9,940, with no advance notice to the recipient of the exact deduction. The sending business cannot predict the net receipt, making reconciliation difficult and creating disputes in supplier relationships. The FX conversion, which typically happens at one point in the correspondent chain (usually the sending bank), adds another layer: spreads of 1.5-3% on common currency pairs are standard, reaching 3-5% on less liquid corridors. On a $50,000 supplier payment, that FX margin alone represents $750-2,500 in friction cost.

De-risking has made this worse over the past decade. Facing rising AML compliance costs and regulatory penalties, major correspondent banks have systematically exited less-profitable correspondent relationships — particularly in developing markets, smaller island economies, and jurisdictions with perceived AML risk. The number of active correspondent banking relationships globally declined by approximately 22% between 2011 and 2022 according to BIS data. Fewer correspondent relationships means more hops for payments in affected corridors, higher costs, and slower settlement.

Why SWIFT MT Was Inadequate — and What ISO 20022 Actually Fixes

SWIFT’s original MT (Message Type) messaging format, designed in the 1970s, was not built for the data richness that modern compliance and reconciliation workflows require. An MT103 single customer credit transfer message carries a maximum of approximately 140 characters of remittance information in unstructured free-text fields. That is not enough to carry an invoice number, PO reference, tax identifier, and payment purpose simultaneously in a structured format that receiving systems can process automatically. The result: manual keying of payment details at the receiving end, high rates of payment-detail mismatches, and slow reconciliation cycles.

ISO 20022, the data-rich XML-based messaging standard that SWIFT’s cross-border network is migrating to through 2025, is a genuine improvement on this dimension. Its pacs.008 message for customer credit transfers can carry structured remittance information, multiple invoice references, tax IDs, purpose codes, and legal entity identifiers (LEIs) in machine-readable fields. For treasury and finance operations doing high-volume international payments, the ability to auto-reconcile against open invoices using structured data embedded in the payment message is a meaningful operational improvement.

What ISO 20022 does not fix is speed or cost. The migration changes the data format carried through the correspondent banking network; it does not change the network topology or the number of intermediary hops. Settlement timing is determined by the correspondent banking chain and the schedules of the institutions in it, not by the message format. FX spreads are determined by liquidity and commercial margin decisions, not by message standards. ISO 20022 is an infrastructure upgrade that will produce real operational efficiency gains over a 5-10 year horizon as systems are updated and straight-through processing rates improve. It is not a solution to the fundamental economics of correspondent banking.

Payment-on-Behalf-Of Models and the New Intermediary Layer

The most pragmatic near-term improvements in B2B cross-border payments have come not from reforming the correspondent banking network but from building around it. A category of specialist B2B payment platforms — Airwallex, Wise Business, Currencycloud (now part of Visa), and others — has achieved meaningfully better pricing and speed by pre-positioning liquidity in local markets rather than routing through correspondent chains.

The model works roughly as follows: the platform maintains local bank accounts or electronic money institution accounts in each supported market. When a customer initiates a cross-border payment, the platform debits the customer’s source account, converts the currency using its own FX position (at interbank-proximate rates), and credits its local account in the destination market. From there, a domestic payment — which is fast, cheap, and final — reaches the end recipient. The correspondent chain is bypassed entirely for the cross-border leg; instead, the platform’s own balance sheet and treasury operations absorb the currency mismatch and settlement risk.

Airwallex processes billions in annualized payment volume using this model, has achieved licensed or regulated status in over 50 jurisdictions, and offers FX spreads of 0.2-0.5% on major currency pairs to business customers — dramatically lower than the 1.5-3% typical of bank FX. Wise Business achieves similar economics. Currencycloud, integrated into Visa’s network post-acquisition, brings this model to bank and fintech partners that want to offer competitive cross-border payments without building their own multi-currency infrastructure.

The limitation of this model is corridor coverage. Maintaining pre-positioned liquidity in 50+ markets requires capital, regulatory licenses, and local banking relationships. Coverage on major corridors (USD, EUR, GBP, SGD, HKD, AUD) is excellent. Coverage in less liquid markets — West African currencies, some Central Asian currencies, certain Pacific Island jurisdictions — is thinner or non-existent, and these corridors often represent the highest commercial need.

The B2B cross-border payments market is not broken everywhere — it is broken unevenly, with the worst friction concentrated in corridors with less liquidity, weaker correspondent banking coverage, and less regulatory standardization. Solutions that work well for a Singapore technology company paying a US SaaS vendor perform much less well for a Nigerian importer paying a Chinese manufacturer. That uneven friction map is where the next generation of cross-border payment infrastructure needs to focus, and where the most interesting fintech investment theses are currently being written.