Visa sketches a hybrid payments stack where stablecoins handle AI-era micropayments

Visa is drawing a sharper line between the payments humans make and the payments software is likely to make on its own. In new research with Artemis, the company argues that the rise of agentic commerce will split the market into two operating modes: conventional consumer transactions that still fit neatly on card networks, and machine-native micropayments that are better matched to stablecoin settlement. That framing matters because it moves the stablecoin conversation away from trading venues and toward a payments use case with a clear economic rationale. If AI agents begin purchasing data, compute, API calls and other digital services at high frequency, the underlying rail has to support tiny values without turning fees into the dominant cost.

The report breaks agentic commerce into macro and micro flows. Macro commerce covers the familiar cases where an agent acts for a user in a human-scale purchase, such as booking travel, paying a subscription or completing a retail checkout. Micro commerce is different. It describes software paying other software in repeated, low-value increments, often well below one dollar, and sometimes for fractions of a cent at a time. That distinction is important because the economics of the two categories diverge quickly. A card authorization model that works for a hotel booking does not map cleanly onto thousands of autonomous API payments made every hour.

Visa’s own data review suggests that this is no longer a hypothetical design exercise. In its July 14 thought-leadership note, the company said x402, an internet-native payment standard incubated by Coinbase and Cloudflare and now under the Linux Foundation, had processed roughly $15 million in adjusted volume across about 109.6 million transactions since launch. The same Visa note said the newer Machine Payments Protocol, built by Stripe and Tempo with Visa contributions, had already handled about $25,000 across roughly 115,000 transactions in its first weeks. Separately, the x402 Foundation now reports tens of millions of monthly transactions and more than $24 million in recent volume on its public site, which reinforces the direction of travel even if the reporting windows differ.

The deeper point is cost structure. Stablecoin payments on newer blockchain rails can settle for fractions of a cent, which makes sub-dollar commerce feasible in a way that fixed card fees historically did not. Visa’s report also points to account abstraction, passkeys and sponsored transactions as supporting infrastructure that can hide blockchain complexity from end users while keeping the settlement leg efficient enough for machine-speed activity. In practice, that means the winning architecture may not look like a pure crypto stack or a pure card stack. It is more likely to be an orchestration layer where software chooses the most suitable rail for each leg of a task.

That is also why Visa is not presenting stablecoins as a replacement for cards. The company’s message is that cards remain well suited to proxy purchases and larger merchant-network transactions, while stablecoins fit the machine-native payment loops that AI systems will increasingly need. Visa says the standards environment is already converging in that direction. Its report points to card-side trust and authorization frameworks such as Trusted Agent Protocol, Agent Payments Protocol and Visa Intelligent Commerce, while also noting that crypto-native systems are increasingly adding the controls and identity layers required for enterprise use. The end state looks less like a winner-take-all contest and more like a blended stack with different trust assumptions at different layers.

For the RWA and stablecoin market, that is a meaningful shift in narrative. Stablecoins have already found product-market fit in crypto trading, treasury movement and cross-border transfers, but agentic commerce would create a new class of recurring operating demand. If AI agents start paying for cloud workloads, model inference, data feeds and software tools, the stablecoin leg becomes a working rail for commerce rather than just a parked cash instrument. That, in turn, raises the value of issuer reliability, redemption plumbing, transaction observability and policy controls, all of which matter more in institutional settings than raw token velocity alone.

The hard part is not only settlement. Visa’s report makes clear that delegated authority, fraud handling, reversibility and liability are still unresolved when an autonomous system initiates the transaction. Those questions will decide how quickly agentic payments move from experiments into production. But the strategic signal is already visible: one of the world’s largest payment networks is now openly describing stablecoins as the right tool for a class of commercial flows that card economics struggle to serve. For RWA builders, that is a stronger and more durable signal than another short-lived crypto payments pilot.