The world of payments for AI agents is rapidly dividing into two camps. On one side are traditional cards for large purchases made on behalf of humans; on the other, stablecoins for microtransactions between programs. This is the conclusion reached by experts from the payment giant Visa in collaboration with the blockchain analytics company Artemis.
Two Categories of Agent Payments
Analysts have identified two fundamentally different types of operations:
- Purchases on behalf of humans: booking tickets, managing subscriptions, ordering goods. These transactions are essentially no different from regular online shopping and can be processed through card infrastructure.
- Micropayments between AI agents: transfers of amounts under $1 for individual API requests, access to data, computing power, or the use of software tools.
It is the second category that is becoming the main driver for the adoption of stablecoins. The fixed fees of card networks make them completely unprofitable for such operations. At the same time, on the Base and Solana networks, the cost of a transaction is fractions of a cent.
Protocols of the Future: x402 and MPP
The report examines two open payment protocols. x402, whose main activity is concentrated on Base, Solana, and Polygon, and the Machine Payment Protocol (MPP), which supports settlements both through blockchains and through traditional payment systems.
Data from Artemis as of April 21 shows impressive volumes. After excluding fictitious and test operations, approximately $15 million was processed through x402 across 109.6 million transactions. For MPP, the figures are more modest: about $25,000 and 115,000 transfers. However, the raw numbers are significantly higher — $135.7 million and 178.3 million operations via x402, highlighting the scale of the data verification problem.
Hybrid Model and Unresolved Issues
The authors predict a hybrid model where cards will be used for authorization and purchases on behalf of humans, while stablecoins will be used for automatic settlements between programs. Both payment methods could be applied at different stages of executing a single task.
The main unresolved problem is trust. It is not yet clear who bears 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 for human actions, and a mechanism for disputing thousands of automatic transactions per hour simply does not exist. An additional risk is posed by malicious instructions that can redirect a payment or force an AI agent to make an unplanned purchase.
Expert Opinion: The market for agent payments is in its infancy, but its potential is colossal. Stablecoins are indeed becoming the ideal tool for microtransactions, where card infrastructure fails. However, without clear rules of responsibility and dispute mechanisms, the mass adoption of such payments will remain in question. Those who solve this problem first will gain a key advantage in the new economy of AI agents.