Stablecoin infrastructure and traditional payment cards will occupy different niches in the ecosystem of payments made by AI agents. This conclusion was reached by experts from payment giant Visa, together with blockchain analytics company Artemis, following an in-depth study of the agent payment market.
Analysts identified two fundamentally different types of transactions performed by artificial intelligence. The first category is purchases made on behalf of a human: booking tickets, managing subscriptions, ordering goods. These operations are essentially similar to regular online commerce and can be serviced through traditional card systems. The second category is frequent transfers of amounts less than one dollar between programs. This refers to paying for individual requests to online services, accessing data, renting computing power, or using software tools.
This is precisely where, in my assessment, the key tipping point lies. The fixed fees charged by card networks make them unsuitable for such microtransactions. At the same time, on the Base and Solana networks, transaction fees amount to fractions of a cent, opening up an entirely new economic landscape for interaction between AI agents.
The report examines in detail two open payment protocols already operating in this field: x402, whose main activity is concentrated on Base, Solana, and Polygon, as well as the Machine Payment Protocol (MPP), which supports settlements both via blockchains and through traditional payment systems.
Real Volumes and the Problem of Data Reliability
According to Artemis data as of April 21, after excluding identified fictitious and test operations, approximately $15 million passed through the x402 protocol across 109.6 million transactions. For MPP, the figures were significantly more modest — roughly $25,000 and 115,000 transfers, respectively. It is important to note that the original unfiltered data was noticeably higher: $135.7 million and 178.3 million operations via x402, as well as over $38,000 and about 184,600 transactions via MPP. The difference is due to the exclusion of operations that Artemis classified as fictitious activity or testing.
The study's authors envision a hybrid model for the future, where cards will be used for authorization and large purchases on behalf of humans, while stablecoins will be used for settlements between programs. Both payment methods could be applied at different stages of executing a single task.
The Unresolved Problem of Trust
Experts cite trust as the main obstacle to the mass adoption of such payments. It remains unclear who should bear responsibility for an erroneous or unauthorized purchase: the owner of the AI agent, the platform operator, the model developer, or the seller. Existing payment reversal rules are designed exclusively for transactions made by humans. An established mechanism for disputing thousands of automatic transactions per hour does not yet exist. An additional risk is posed by malicious instructions capable of redirecting a payment or forcing the AI agent to make an unplanned purchase.
My expert assessment: the agent payment market is in its infancy, but stablecoins have a clear structural advantage for microtransactions. However, without resolving the issue of legal liability and creating mechanisms for disputing transactions, this segment risks remaining a niche experimental field. Integration with traditional card systems through hybrid models is the most pragmatic path to scaling.