88% of AI Capital Lands in the US. The Best Infrastructure for Agentic Finance Is Somewhere Else.

In the first quarter of 2026, AI-related startups raised $297 billion globally. Nearly 88% of it went to US-headquartered companies. The rest of the world split the remaining $35 billion across Europe, Asia, and LatAm. Brazil and Mexico, together the dominant force in regional venture funding, attracted under $300 million in AI startup capital across all of 2025. That capital allocation is not a reflection of where the best AI opportunities are. It is a reflection of where the money is currently paying attention.

Agentic finance — AI systems that make autonomous credit decisions, execute compliance workflows, flag fraud in real time, and manage treasury operations without human authorization — requires three things: reliable data access, structured financial APIs, and a large population whose financial behavior exists in digital form. Brazil has all three at a level that no developed market has matched. Pix processes more than 70 million transactions per day and has been live since 2020. Open Finance APIs give registered institutions access to customer financial data across all connected institutions with consent, through live infrastructure that has been operating since 2023. The Drex tokenization layer, now redesigned from a digital currency framework to a programmable settlement system for tokenized assets, adds a programmable reconciliation layer over which financial contracts can execute automatically. The Banco Central has built the technical preconditions for agentic finance faster than any other central bank in the world.

The gap between available infrastructure and actual AI application on top of it is striking. The 20% of credit operations currently affected by lien duplications — a specific, quantifiable inefficiency that Drex's reconciliation layer is designed to eliminate — represents an immediate product opportunity for AI-native credit platforms once the settlement layer is live. The same data infrastructure that allows Pix to process a payment in seconds allows an AI underwriting agent to verify income, assess credit history across institutions, and approve a loan in a comparable window. No US bank has the data architecture to do that today. Brazil built it by regulatory mandate, which means every financial institution in the market operates on the same interoperable rails.

The capital gap has a structural explanation: US venture firms have US LPs who want US returns, and the regulatory and tax complexity of investing in Brazil has historically been a friction that requires local presence or a specific thesis. What is changing is the thesis. As AI-native financial applications require exactly the kind of structured data infrastructure Brazil has already built, and as Brazilian fintech companies demonstrate that they can compete on unit economics rather than user acquisition, the threshold for global capital is shifting. Itaú Ventures launched with R$500 million in committed capital focused on this intersection. Founders Fund committed $100 million to Brazil in 2026 despite — or because of — a 67% year-over-year decline in Brazilian venture activity. The conviction is that the infrastructure is real and the capital will follow.

The argument that Brazil is structurally disadvantaged by high interest rates, political volatility, and limited exit options is real. The Selic at 14.5% creates a hurdle rate that punishes early-stage companies and selects hard on unit economics. But it also means the companies that survive to scale have done so against a harder benchmark than their counterparts in zero-rate environments. The most durable AI finance applications will not be built in markets where data infrastructure is still being constructed. They will be built in markets where the infrastructure is already live, the consumer population is already digital, and the specific financial inefficiencies are measurable enough to build a product around. Those conditions are rare. Brazil has them. The capital hasn't caught up yet — which is either a problem or an opportunity, depending entirely on your time horizon.