The Silent Shift: Why the Ground Underneath Fintech Is Moving
There is a quiet shift happening in fintech right now. It is a shift in what fintech competes on, and more importantly, who controls the infrastructure it runs on.
For years, venture dollars went primarily into features layered on top of banking systems owned by others. The layer underneath was static. So if businesses wanted to differentiate, the place to compete was on the surface: better UX, more payment rails, cheaper processing, faster onboarding.
That era is ending. We’re now seeing movement with the layer underneath, and my concern is that most fintech roadmaps have not caught up.
Features are getting cheap. Primitives are not.
Consider what’s emerging at the infrastructure layer: programmable money, tokenized credit, accounts that software — not just humans — can hold, compliance written as code instead of policy. These are primitives, not features.
The distinction matters. A feature is a thing your product does. A primitive is a thing other products are built on. Once you own the primitive, the features build themselves on top of it.
I keep hearing people talk about product differentiation at the UX layer. That's fine, but it’s short-sighted and it won't be the sorting function. The companies that win the next decade won’t be the ones with the prettiest checkout. They’ll be the ones who either own the primitives or chose the right partner to build on top of. This is a strategic decision, and it compounds in both directions.
Three forces just converged
Three things just happened at the same time. Any one of them opens a window. All three at once is a door — and doors close.
Assets are moving onchain. Stablecoin market cap is now above $300 billion [1]—large enough to sit in the same order of magnitude as meaningful parts of the regional banking system. More importantly, this is no longer speculative capital. It is increasingly being used for treasury management, vendor payments, and working capital. In other words, real commercial activity is starting to move onto new rails. The strategic question is who will control that infrastructure.
Software is starting to transact on its own. Legacy payment systems were designed for human workflows, operating hours, and known counterparties. They are poorly suited to autonomous agents moving funds in real time, particularly in environments that require always-on availability, programmable controls, and immediate settlement. Businesses that start building for that world now will be in a much stronger position than those trying to retrofit their infrastructure once the volume is already there.
Regulation is arriving. The GENIUS Act and stablecoin rulemaking expected in 2026 and 2027 should give regulated issuers a clearer defined operating perimeter than the market has had in years. As the ambiguity recedes, the advantage shifts toward firms that are already building within a regulated framework.
Taken together, these forces matter. Demand is growing, regulation is becoming clearer, and a new class of software-driven payment activity is emerging that legacy rails were not built to support. This combination only comes around every couple of decades. It’s the reason we’ve prioritized infrastructure-level decisions over product-level ones.
What we decided to build and why
Most fintech infrastructure today is held together with reconciliation. There’s a wallet on one side, a ledger on the other, and a section in the middle that an engineering team rewrites every few months. If you look closely, you’ll see that a lot of fintech roadmap in 2026 is reconciliation work.
The strategic decision we made at Cross River was to unify the layer underneath before the market demanded it: one core where programmable balances, identity, compliance, and credit coexist as shared infrastructure.
When the layer underneath gets unified, a lot of engineering effort stops being necessary. The work your team has been spending on the primitive layer becomes work you can spend on what differentiates your product. Which, presumably, is the reason you started the company.
What this changes in practice
Below are three concrete examples, each of which is a real commercial conversation we are having right now.
Payments. A global marketplace paying out to ten countries today typically prefunds ten corridors, KYCs each recipient twice, and reconciles two ledgers nightly. Working capital sits in floats it can’t touch. With unified rails, one balance funds every corridor on demand. KYC happens once. Settlement is final. The payout is the same with less capital tied up. Cross-border stops being a treasury problem and becomes a product capability. Businesses that were priced out of global payout by float costs can finally offer it.
Credit. A card issuer financing receivables today originates, warehouses, waits weeks for the draw, and renegotiates covenants once a year. With tokenized receivables and onchain capital pools, each receivable can be sold at origination into a vault with pre-committed capital. Capital turns in hours. Multiple investors compete on price. Covenants are encoded in the vault and checked every block. Cost of capital can come down 100 to 200 basis points. Private credit distribution starts behaving like a capital market, where cost of capital is driven by underwriting performance rather than banking relationships. That is a different industry structure.
Agentic payments. This is where the gap between conference demos and production reality is widest. Wallet rails were built for humans. They don’t have the controls a CFO wants, the audit trail a regulator requires, or the recourse a corporate counterparty needs. With the right primitives in place, agents get a real account. Spending policy is enforced by the ledger. Compliance applies at every transaction. There is a regulated counterparty on the other side of every flow. Agentic commerce becomes something you can put in front of a CFO, an auditor, and a regulator at the same time. The businesses that will ship this in 2027 are the ones who stopped trying to make wallet rails do something they weren’t built for.
Three questions to ask your infrastructure partner
Across banks, fintechs, crypto platforms, and even the new wave of stablecoin-as-a-service companies, a lot of providers will tell you they can deliver this. The difference shows up when something goes sideways — or when you try to extend the relationship.
1. Regulatory ownership. When a regulator calls about a transaction on your platform, whose name is on the response? If the answer is “we’ll help you figure it out,” that’s a vendor, not a partner. There’s a meaningful difference, and you feel it the first time something goes sideways.
2. Composability. If you want to extend credit against onchain balances next year, what do you rebuild? If the answer involves a new vendor, a new integration, and a new compliance review, you don’t have a platform. You have a series of one-off products with a shared logo.
3.Readiness for what’s next. What happens to your compliance model when half your flows are driven by AI agents? This is the tell. Most providers haven’t thought about it seriously, and the ones who have are the ones building the infrastructure now. You don’t want to have discussions about retrofitting when the volume becomes real (which could be as soon as next year).
The direction is not in doubt. Only the timing.
In 2028, “fintech” and “crypto” won’t be separate categories. Infrastructure, accounts, and compliance will run on the same core. The only real question will be who builds it and who rents it.
The businesses that win will be the ones making infrastructure decisions today as if that’s already true.
Build on the primitives. The features will follow.
Cross River processes roughly $1T in payments annually, has originated more than $100B in loans, and powers Coinbase, Affirm, Stripe, and others. We are building the regulated onchain infrastructure layer where programmable balances, credit, identity, and compliance sit on the same core — so builders compete on their product, not on plumbing. If the three questions above gave you pause about your current setup, we should talk.
[1] Top Stablecoin Tokens by Market Capitalization. CoinMarketCap. Accessed May 12,2026.
