The end of correspondent banking (or at least the interesting part)

Correspondent banking is not dying. There will still be a nostro account in Frankfurt when my kids are in university. What is happening is stranger, and more important: the model is losing the parts of its business that ever earned interesting margins, and holding onto the parts that do not.

I have watched this from a distance for years and up close for about two. The pattern is clear now.

The model was always a trust supply chain

Strip away the terminology and correspondent banking is a chain of favors. Bank A holds an account at Bank B in a country where A does not operate. B does the same at C. When a customer at A wants to pay someone in C's country, the message travels the chain, and each bank takes a piece for the risk of having its own capital exposed to the bank next door.

That works. It has worked since the Medici. The problem was never that it does not function. It is that every layer of the chain got paid to solve a coordination problem that computers are now solving on their own.

What the last five years actually changed

Three things happened, roughly at once, and the industry has not fully caught up to what their combination means.

First, USD stablecoins reached daily volume where they stopped being novel and started being infrastructure. Not because anyone declared them so, but because businesses started paying suppliers with them and noticed the money arrived on Sunday.

Second, regulators in the jurisdictions that matter for global finance stopped treating "stablecoin" as a hostile category and started writing rules for it. MiCA in the EU, stablecoin bills in the US, similar movements in Singapore and Hong Kong. The legal ambiguity that was a moat for the correspondent network became a moat for regulated non-bank players.

Third, treasurers got tired. This one is underrated. Sit across from someone running FX for a mid-sized importer and they do not want a lecture on the future of money. They want to know why they still pay 180 basis points on a EUR to NGN corridor and why their supplier receives the money four days after they sent it.

What actually survives

Correspondent banking does not go away. It retreats to the corner of the market where its structural advantages still matter: very large tickets, jurisdictions where regulated non-bank rails do not yet exist, situations where the customer is a bank itself and counterparty risk is the whole point of the trade.

Everything else migrates. Not fast, and not cleanly. But it migrates.

Corporate payables to countries with mature stablecoin off-ramps? Migrated. Remittance? Mostly migrated on the front end, with the back end following. Supplier payments in emerging markets where the recipient has been struggling to hold dollars locally anyway? That is the corridor changing right now.

The awkward middle

What makes this transition weird is that most banks are still in the correspondent business but sell their retail customers a "modern" cross-border experience by routing through a fintech partner behind the scenes. The customer sees "settled in minutes." The bank sees a smaller margin. The correspondent chain sees a bit less flow.

That is a stable equilibrium for a while. It is also how these transitions usually happen. Nobody announces the shift. They quietly re-route flow and let the old system carry the residual volume that is not worth changing.

What we are watching for

A few tells for how the rest of this decade plays out.

If large banks start ripping and replacing nostro accounts in specific corridors rather than layering new rails on top, that is the loud phase. Right now they are mostly layering.

If SWIFT volume starts flattening in specific corridors, not overall but corridor-level, that is the quiet indicator. Overall SWIFT volume can grow while its share of any given corridor's economics shrinks. That is already true in a handful of routes.

If a stablecoin-native settlement layer becomes the default for a specific new fintech vertical (cross-border payroll, B2B marketplace payouts), that is the shape of what comes next. The old network will still be there. It will just be less of the interesting part.

I will stop short of predicting when. I have been wrong on timelines before. But if you are building anything that touches cross-border money right now, the assumption I would operate under is this: the correspondent model is your fallback, not your default. That is a genuine reversal from ten years ago, and it happened without a headline.

---

Resrv Technologies provides regulated stablecoin settlement infrastructure for institutions moving value across borders. To learn more about Nexus settlement accounts or the RDAN network, get in touch.