AI Payments: The $1.9 Trillion Flow Engine

The existing payment flow that AI agents are tapping into is massive. Businesses running on Stripe generated $1.9 trillion in total volume last year, a 34% year-over-year increase that equates to roughly 1.6% of global GDP. This scale represents the foundational engine AI-driven commerce must plug into.

The shift from experimentation to mainstream adoption is now underway. Visa has already seen hundreds of secure, agent-initiated transactions completed in real-world pilots, signaling the technology is moving beyond the lab. This activity is supported by a growing ecosystem of partners actively building the infrastructure.

The core revenue model is straightforward: transaction fees on this existing flow. New services like Stripe's machine payments layer on top, charging for API calls to facilitate these automated transactions. The engine is operational, and the fee structure is clear.

The Liquidity Layer: Network Tokens and Fraud Controls

The critical payment primitive enabling this flow is Stripe's Shared Payment Tokens (SPTs). This system lets AI agents initiate payments with a customer's permission and preferred method, without exposing underlying card details. It's now expanding to support network tokens like Visa and Mastercard, creating a secure, scalable credential layer.

Visa's edge-based behavioral intelligence is the essential fraud control that gives merchants confidence. Its Visa Intelligent Commerce initiative, powered by AI, provides the security moat needed for real-world adoption. This capability is already being tested in hundreds of live transactions, proving the system can handle agent-driven purchases securely.

The infrastructure is broadening to include flexible financing. SPT support is being extended to BNPL methods like Affirm and Klarna, making agentic payments accessible to a wider customer base. This integration increases potential transaction volume and aligns with the growing $300 billion BNPL market.

Catalysts and Risks: Flow Acceleration vs. Flow Disruption

The next major catalyst is the expansion of the Shared Payment Token (SPT) infrastructure to more payment methods. Stripe is rolling out support for network-led agentic payments with Mastercard and Visa, alongside BNPL tokens. This makes Stripe the first provider to unify these credentials under a single primitive, dramatically lowering the integration barrier for sellers and unlocking access to a broader customer base.

The primary risk is the fraud and abuse potential from manipulated AI agents. As the flow scales, the system must manage the threat of compromised or malicious agents initiating unauthorized transactions. This could increase chargebacks and erode merchant trust, making the robust fraud controls from partners like Visa Visa Intelligent Commerce a critical, non-negotiable layer.

The ultimate market size is global payment revenue, which grew 7% annually to over $2.4 trillion from 2018–2023. This provides a long-term TAM for the new flow of agentic commerce, but the near-term path depends on successfully navigating the expansion of payment methods and the persistent challenge of securing the automated payment layer.