Every institution that wants to move value across a border needs a relationship. That is the oldest rule of correspondent banking, and it is also why cross-border payments remain slow, opaque and expensive. The problem is not effort or willingness. It is arithmetic, and arithmetic does not negotiate.

The mathematics of relationships

A bilateral model connects two parties directly. Connect ten institutions that way and you need dozens of arrangements; connect thousands and the count runs into the hundreds of thousands. This is the N² problem: every new participant must be connected to every existing one, and each connection carries its own documentation, limits, prefunding and operational overhead.

In practice, no network ever reaches a full mesh. Banks build links where the flow justifies them, and the map stays incomplete. Adding one more relationship does not fix that; it adds one more node to an already partial graph.

What an incomplete graph costs

The number of active correspondent banks worldwide fell about 22% between 2011 and 2019 while payment volumes rose. In Latin America, active relationships fell about 30% since 2012. Fewer links carrying more volume means more concentration, more hops and less competitive pressure on price.

For the client at the end of the chain, that appears as a payment routed through several banks before it arrives.

From relationships to a connection point

The alternative is architectural rather than commercial. Instead of every institution opening accounts everywhere, a single hub holds correspondent accounts for institutions — reverse correspondence. Participants keep one connection rather than building a bilateral relationship for every corridor.

When flow passes through one point, it can be validated once and netted. If bank A sends €10m while participants B and C receive €8.5m, only the €1.5m difference leaves the hub. The remainder settles internally.

SUPA is designed as that neutral layer: infrastructure and a protocol for cross-border movement of value, not a competitor to the institutions that use it. Clients stay clients of their own bank or fintech.

In short

  • Bilateral relationships scale quadratically; a shared connection point does not.
  • Correspondent networks have been shrinking even as volumes grow.
  • A hub can validate once and net offsetting flows internally.
  • Neutrality is the operating principle: the layer connects institutions without competing for their clients.

Discover how participation works at /institutions/.