Money, rewritten in software

Fintech's first act was apps on top of old rails. The second act rebuilds the rails themselves.

The first decade of fintech was, mostly, a beautiful interface on a creaking back end. Better onboarding, cleaner apps, faster cards: all running on settlement systems and credit models that predate the internet. It was a real business, and a temporary one. The neobanks taught a generation to expect their money to behave like software. They did not, for the most part, change what the money was doing underneath.

The second act is structural. Money itself is becoming programmable; finance is dissolving into infrastructure that other companies embed. Underwriting is becoming a machine-learning problem rather than a committee. And the rails that move value, across borders, across currencies, increasingly across software agents transacting on our behalf, are being rebuilt from the protocol up.

The tell is where the talent and the margin are migrating. The interesting companies have stopped trying to be the bank a consumer chooses and started trying to be the financial capability a thousand other products depend on. That is a quieter ambition and a larger one. A consumer brand competes for attention and churns; an infrastructure layer competes for integration and compounds. Once a payout system, a ledger, or a compliance engine is wired into someone else's product, it stops being a vendor and starts being load-bearing.

From apps to infrastructure

The shift from apps to infrastructure changes the unit of competition. In the app era, the moat was brand and acquisition cost, and both eroded the moment a better-funded entrant bought the same ad inventory. In the infrastructure era, the moat is the cost of ripping you out. Financial infrastructure embeds itself in the parts of a business that no one wants to touch twice: money movement, reconciliation, regulatory reporting, the ledger of record. That stickiness is real, but it is earned slowly and only by companies that are trusted to be correct, every time, at scale.

Correctness is the whole game, and it is why this market resists the move-fast instincts that work elsewhere in software. A consumer app can ship a bug and patch it by morning. A payments rail that loses a reconciliation, mishandles a regulated flow, or settles to the wrong account does not get a patch; it gets a regulator and a withdrawn license. The companies that win infrastructure are the ones that treat boring reliability as a feature and build for the audit before anyone asks for it. We underwrite that temperament as carefully as we underwrite the technology.

Why now, and why everywhere

This rebuild is not a developed-market story. The largest gains accrue where the legacy system is weakest and the population is youngest and most mobile-native. The same infrastructure thesis that produces a US embedded-finance leader produces a market-defining payments company in South Asia or Latin America, often faster, because there is less to unwind.

The leapfrog is not theoretical. In markets without entrenched card networks or universal bank branches, value moved straight to mobile, and the rails being built there are native to how money actually circulates: small, frequent, instant, mobile-first. Real-time payment systems went from novelty to default in a handful of years in several of these markets, and the companies that built on top of them did not have to coax anyone off a legacy habit. They simply met a young, formalising economy where it already was. That is a structural tailwind a mature market cannot manufacture.

There is a second-order effect we watch closely. As formal financial rails reach people who were previously invisible to the system, they generate the data that makes those same people creditworthy for the first time. A payment history becomes an underwriting signal; a merchant's transaction flow becomes the basis for a working-capital loan that no committee would ever have approved. Infrastructure that brings the next participant into the formal economy does not just serve a market; it expands one. That compounding loop, more rails, more data, more access, more rails, is among the most attractive dynamics we underwrite anywhere.

We back the companies re-plumbing finance, for businesses, and for the next billion participants entering the formal economy.

Where the two theses meet

When intelligence becomes cheap and money becomes programmable, the two theses converge: AI-native financial infrastructure. That intersection is where we spend a disproportionate share of our conviction.

The convergence is concrete, not slogan. Underwriting, fraud detection, and compliance are pattern-recognition problems drowning in data, exactly the kind of work that abundant intelligence is built to do, and exactly the kind of work where a wrong answer is expensive enough that the system has to be trusted, explainable, and auditable. The companies that can fuse a defensible data advantage with the operational discipline of regulated finance will own functions that have resisted automation for decades. And the arrival of software agents that transact autonomously will demand a new layer of financial plumbing: identity, authorisation, limits, and settlement designed for machines as first-class economic actors rather than as scripts borrowing a human's credentials.

Our posture here is to back operators who have lived inside regulated finance and understand that the constraint is rarely the model; it is the institution's willingness to trust it. We look for teams who can win the trust of a counterparty's risk and compliance functions, because that, far more than the demo, is what gets infrastructure adopted. Money is being rewritten in software. We intend to fund the people writing the parts that have to be right.