Remittance infrastructure: what sits under the apps

Every remittance app sits on infrastructure the sender never sees, and that infrastructure, not the app, decides what the transfer costs and how long it takes. Remittances are one of the largest cross-border flows in the world: the World Bank expected flows to low- and middle-income countries to reach $685 billion in 2024, larger than foreign direct investment and official development assistance combined. Yet sending $200 still costs 6.36% on average, and 8.78% in Sub-Saharan Africa. To understand why, you have to look under the apps.

Layer one: the licensed originator

The brand the sender touches is a regulated entity in the sending country: a money transmitter in the United States, an e-money or payment institution in Europe and the UK, or a bank. This layer owns the customer, the sending-side compliance obligations, and the collection of funds. It rarely owns the journey. Once the money is collected, the originator hands the payment to whatever infrastructure reaches the destination.

Layer two: aggregator networks

Between the originator and the destination sit the aggregators: networks that maintain direct connections into local payout systems so that individual providers do not have to build them corridor by corridor. Thunes reports reach into 140+ countries across 90+ currencies through a single connection; Nium and Ripple operate on the same principle, with Ripple reporting payout coverage in 60+ markets on a network it says has processed over $100 billion, settling in fiat or in regulated stablecoins including its own RLUSD. These networks are the reason a mid-sized remittance brand can offer 80 destination countries without 80 licenses and 80 banking relationships.

The aggregator model works, and it has limits. Each network covers the corridors it covers. Coverage is a commercial asset, so networks overlap in the busy corridors and thin out exactly where correspondent coverage is also thinnest. And an aggregator is one more intermediary taking a margin on every payment that crosses it.

Layer three: the correspondent fallback

Where no network reaches, the payment falls back to the correspondent chain: a sequence of banks holding nostro and vostro accounts for one another, each adding time, fees, and cut-off risk. This fallback is shrinking. The BIS counted a decline of about 22% in active correspondent relationships between 2011 and 2019, and the retreat has hit remittance-heavy regions hardest, a dynamic we cover in de-risking. The corridors most dependent on the fallback are the corridors where it is disappearing.

Layer four: local payout rails

The last mile is local: a credit to a bank account over the domestic payment system, a mobile money wallet, or cash over a physical agent network. This layer sets the floor on speed (a payout is only as fast as the local rail) and much of the floor on cost in cash-heavy markets, where agent commissions are a real expense no upstream efficiency removes.

The floor is structural

Put the four layers together and the pricing puzzle resolves. A remittance crosses a licensed originator, usually an aggregator, sometimes a correspondent chain, and a local payout rail, and each layer prices in its costs: compliance, pre-funding in destination currencies, FX conversion, and margin. The proof that the floor is structural is what happens when a provider replaces the layers with direct integrations: Wise, which connects directly to domestic payment systems, reported an average cross-border take rate of 0.52% in its 2026 fiscal year, roughly a twelfth of the global average price. The gap between 0.52% and 6.36% is the infrastructure.

That is also why the G20 target, a 3% global average by 2030 with no corridor above 5%, has proven so hard to reach. The FSB’s October 2025 assessment of the broader cross-border roadmap concluded that the 2027 targets are unlikely to be met at the global level. Apps have improved for a decade; the layers beneath them have improved far more slowly.

Where stablecoin corridors slot in

Stablecoin rails enter this stack in the middle. In place of the aggregator hop or the correspondent fallback, value moves between the two markets as a regulated stablecoin in minutes, with settlement finality on a shared ledger and no pre-funded account chain. The pattern, fiat in, stablecoin across, fiat out, is common enough to have a name, and we examine it in the stablecoin sandwich and the corridors where it wins in emerging-market corridors. What stablecoins compress is the middle leg. The first and last mile, licensed collection and local payout, remain, which is why ramp coverage and local liquidity now decide where the model works.

What would actually move the target

Three changes move remittance economics at the infrastructure level rather than the interface level. First, more direct connections into domestic instant payment systems, the Wise model, extended by projects like Nexus. Second, settlement legs that do not require pre-funding, whether through interlinked instant rails or regulated stablecoin corridors, releasing the capital that today sits idle in destination accounts. Third, compliance that travels with the payment instead of being re-performed at every layer, so that a $200 transfer does not carry four separate screening costs.

Where Frame fits

Frame is the settlement layer for global finance: one integration to orchestrate payments at scale across fiat rails, stablecoins, and tokenized deposits, with compliance enforced on every transaction and settlement in seconds instead of days.

For the institutions that operate remittance infrastructure, and for the banks, payment providers, exchanges, platforms, and enterprises that build on it, Frame’s rail-neutrality maps directly onto the stack above: a payment enters through one integration and routes over whichever leg serves the corridor, an aggregator network, a domestic rail, or a regulated stablecoin corridor, under the policy that governs it. Frame’s Rules Engine evaluates every transaction against those policies inside settlement, so compliance is enforced once, verifiably, rather than re-checked at every layer. The stack under the apps will keep evolving corridor by corridor. A settlement layer that spans the stack lets its operators price each corridor on the best rail available to it.

See how a rail-neutral settlement layer works: the Frame Blueprint.