Something structural is happening inside the global financial system. Not a pilot program, not a proof of concept — a rewiring. From Mastercard's freshly announced AgentPay framework to JPMorgan's deepening investments in AI startups, the institutions that move money at civilizational scale are no longer experimenting with artificial intelligence. They are building on it.
Mastercard's Q4 2025 earnings offered a window into how far that shift has already gone. Net revenue grew 15% year-over-year on a currency-neutral basis, but the more telling number was in Value-Added Services: a 22% increase, with 18% of that coming from organic growth. That VAS expansion — spanning fraud analytics, tokenization, and data intelligence — is where AI lives inside Mastercard's business model. Approximately 40% of all Mastercard transactions are now tokenized, and over 70% are processed through its switched network, up ten percentage points since 2020. These are not incremental improvements. They reflect a payment infrastructure that has been quietly rebuilt around machine-readable, AI-optimizable data flows.
The AgentPay framework — Mastercard's initiative to enable AI agents to execute payments autonomously within defined parameters — represents the next layer of that buildout. As agentic AI systems proliferate across enterprise software, the ability to embed compliant, auditable payment execution into those agents becomes a competitive moat. Mastercard is positioning itself as the rails those agents run on.
JPMorgan is approaching the same transformation from the other side of the ledger. Rather than building every capability in-house, the bank has been systematically investing in and partnering with AI-native fintech startups across credit decisioning, compliance automation, and customer intelligence. The bet is that the AI layer of financial services will be populated by specialists — and that owning a stake in that ecosystem is as strategically important as building internal tools.
The broader ecosystem reflects the same logic. eToro's trading platform has outperformed expectations as AI-assisted portfolio tools drive engagement. Coinbase has demonstrated resilient crypto revenue even through market volatility, partly on the strength of its institutional custody and analytics infrastructure. Robinhood continues to expand its product surface with AI-driven personalization.
Meanwhile, the compliance layer is catching up. Companies like FairPlay AI and Cleareye.ai are building the explainability and auditability tools that regulators will require as AI-driven credit and lending decisions scale. The Trump administration's reversal of the H20 chip export ban — easing access to Nvidia's data center hardware for certain markets — has removed a significant constraint on the compute infrastructure that underpins this buildout.
Perhaps the most forward-looking signal is the OP Pohjola–Qutwo quantum-AI research partnership, which is exploring how quantum computing can accelerate AI model training for financial risk assessment. That partnership is years from commercial deployment, but it signals where the frontier institutions are placing long-horizon bets.
What connects these developments is not technology for its own sake. It is the recognition, now widespread among the largest players in global finance, that AI-native infrastructure confers durable competitive advantages in speed, cost, fraud prevention, and customer intelligence. The institutions moving fastest are not doing so because AI is fashionable. They are doing so because the alternative — ceding that infrastructure advantage to faster-moving competitors — is increasingly unacceptable.
The financial system is being rebuilt in real time. The construction is noisy, the standards are still being written, and the regulatory perimeter is shifting. But the direction is no longer in doubt.