Something shifted in cross-border payments over the last few years, and it happened faster than most people expected. The stablecoin layer - once an experiment at the margins - is now a real part of how money moves globally. You can see it in the deal flow, customer conversations, in where regulatory attention is going. Digital assets have crossed the chasm.

But having crossed the chasm doesn't mean you've arrived. It means the question has changed. And the question the industry is now dealing with - mostly implicitly, sometimes explicitly - is whether the architecture that got us this far is the architecture that takes us to the next stage.

The dominant architecture is what people started calling the stablecoin sandwich a year or two back. Fiat converts to stablecoin at one end, value moves across the chain, stablecoin converts back to fiat at the other end. It works. It has unlocked real use cases: hard-currency corridors where someone on one side urgently wants dollars, corporate treasury moving funds across borders at hours that SWIFT can't touch, situations where you can bundle the off-ramp problem into one partner relationship and move on.

I spent time at a very large GSIB early in my career, building systems reconciling trade feeds to SWIFT feeds, sitting inside correspondent banking. Something I learned - and that I keep thinking about now - is that correspondent banking works. It is expensive, slow, and remarkably opaque, but it reliably moves money. The model is: bilateral relationships, pre-funded positions, corridor-by-corridor expansion. You add a corridor when the economics justify it. You maintain the relationships because the alternative is reduced volume.

What I find genuinely interesting about the stablecoin sandwich is that it has reproduced this model almost exactly. Different settlement rails, same underlying structure. If you peel back the UX improvements and the faster settlement times, the correspondent banking architecture is still there. You still need pre-funded positions on both sides of each corridor. You still build out one corridor at a time. Conversion costs don't disappear - they shift around, sometimes to your benefit, sometimes not - but they stay in the system.

The payment CEOs I talk to who have been running stablecoin sandwich operations for a few years are starting to feel this. Three or four corridors looks great on the unit economics. Twenty corridors - which you need if you're serious about global reach - starts to look like a different business. Many stablecoin orchestrators product roadmaps are filled with integrations to support more corridors to increase volume rather than building better products.

The operational overhead, the compliance exposure per corridor, the working capital tied up in pre-funded positions: it compounds. And margins that looked healthy early on start to come under pressure. Many of the largest banks globally have secretly been removing expensive and risky to operate long tail corridors from their correspondence bank offerings by outsourcing this volume to PSPs and stablecoin orchestrators.

The industry is already sensing this, and you can see it in where the interesting product work is happening. What's emerged is a kind of open-faced stablecoin sandwich: one-way fiat conversion instead of two-way. Stablecoin cards are the clearest example - the cardholder's account is stablecoin-native, the conversion only happens at the point of spend into fiat. You've cut the conversion problem in half. One side of the transaction has gone native; the other side is still touching the legacy rails. It's real progress. But it's still a transitional architecture, not the destination.

This isn't a criticism of the people who built this way. The stablecoin sandwich, and its open-faced variant, were the correct call for the environments in which they emerged. When you're trying to get from zero to adoption, you build for backward compatibility. You operate within the systems where the money already lives.

Every technology does this phase. Remember Vonage or paying for a phone number in Skype? VoIP spent years routing calls between traditional phone endpoints before Skype and WhatsApp came along and made those endpoints superfluous for general use. The stablecoin sandwich is how you get a foothold, not how you achieve the end state.

As more issuers enter the market - across banking consortia, fintechs, and sovereign-backed initiatives - the question stops being which stablecoin you're holding and becomes what you can do with it natively. A network of institutions that can settle with each other directly, without rebuilding bilateral corridors for each relationship, compounds differently than a network of corridors does. That's the transition worth building toward.

There's a difference between a transitional architecture and a destination, and the industry is at the point where that distinction matters for the decisions you're making now. The companies worth watching are the ones that ran the stablecoin sandwich where they needed to, understand exactly what it costs them to operate, and are building deliberately toward something different. The regulatory framework is falling into place. The infrastructure question is the one that deserves more honest attention.

‍