The financial services industry is undergoing a structural shift: AI is no longer a layer bolted onto existing systems but is becoming the infrastructure itself. From payment networks to trading platforms and crypto exchanges, the industry's largest players are redesigning core operations around autonomous, AI-driven capabilities.
Mastercard's Q4 2025 earnings offer a window into how this transformation is already generating measurable returns. The company reported 15% net revenue growth year-over-year on a non-GAAP currency-neutral basis, with its Value-Added Services segment — which encompasses AI-powered analytics, fraud detection, and identity verification — growing at 22%, with 19% of that organic. More than 40% of all Mastercard transactions are now tokenized, and over 70% are processed through its switched network, up 10 percentage points since 2020. These numbers reflect an infrastructure that has been quietly rebuilt around machine-readable, programmable transaction flows.
Central to Mastercard's AI strategy is AgentPay, its emerging framework for enabling autonomous AI agents to initiate and settle payments within defined guardrails. The concept addresses one of the most pressing questions in enterprise AI deployment: how do you give an AI agent spending authority without sacrificing compliance or auditability? AgentPay is designed to answer that, embedding policy controls directly into the payment rail rather than relying on downstream approval workflows.
JPMorgan Chase is approaching the same problem from the investment side, backing a constellation of AI startups focused on financial infrastructure. The bank's thesis is that the next generation of financial tooling will be built on foundation models capable of reasoning across regulatory frameworks, risk models, and market data simultaneously — a capability that static rule-based systems cannot replicate.
On the trading and retail investment side, eToro's platform has outperformed benchmarks by leaning into AI-driven portfolio construction and social signal aggregation. Robinhood and Coinbase, meanwhile, are integrating AI into customer-facing advisory features and fraud prevention, with Coinbase demonstrating revenue resilience even through crypto market volatility — a function partly attributed to its AI-enhanced risk management stack.
The governance layer is keeping pace. Firms like FairPlay AI and Cleareye.ai are embedding explainability and bias-detection directly into lending and underwriting pipelines, responding to regulatory pressure for compliant AI in credit decisions. Cresta is deploying large language models in financial services contact centers, automating complex customer interactions while maintaining audit trails for compliance purposes.
At the research frontier, Finland's OP Pohjola bank has partnered with quantum computing firm Qutwo to explore quantum-AI hybrid models for portfolio optimization and risk simulation — a pairing that remains pre-commercial but signals where long-term infrastructure investment is heading.
A significant regulatory tailwind has also emerged: the Trump administration's reversal of the H20 chip export ban clears a path for broader AI compute availability, reducing a key bottleneck for firms building inference infrastructure at scale.
What is becoming clear across these developments is that financial services AI is converging on a common architecture: autonomous agents operating within programmable compliance envelopes, transacting across tokenized rails, with explainability baked into the infrastructure rather than retrofitted afterward. The institutions moving fastest are those treating AI not as a product feature, but as the operating system of their business.