Visa and blockchain data firm Artemis published a joint report this week proposing that card networks and stablecoins operate side by side as AI agents handle increasing payment volumes. The report argues each payment rail should focus on its optimal use case rather than compete directly. The analysis divides AI-driven commerce into macro-commerce for consumer purchases and micro-commerce for small machine-to-machine transfers under one dollar, with card networks handling larger transactions while stablecoins process frequent low-value payments between software programs.
The report, titled "Agentic Payments from the Ground Up," identifies macro-commerce as covering everyday purchases such as travel bookings and subscriptions where AI agents buy on behalf of individuals. Micro-commerce encompasses small, repeated payments under a dollar between programs, including API calls and computing power transactions. Card rails continue to handle larger purchases effectively, but fixed fees make tiny payments too costly to process. Newer blockchains have reduced settlement costs to fractions of a cent, making them better suited for machine transactions.
x402 and Machine Payments Protocol Process Millions in Transactions
Onchain data supports the payment rail division through two software payment systems launched over the past year. The x402 protocol, developed by Coinbase and Cloudflare and now operated by the Linux Foundation, processed approximately $15 million across 109.6 million payments since May 2025, primarily on Base, Solana and Polygon networks. The Machine Payments Protocol, built by Stripe and Tempo with Visa, settled about $25,000 across 115,000 payments in its first weeks. Each transfer sits well under a dollar, amounts too small for fixed card fees to justify economically.
Regulatory Framework Gaps Remain Unresolved for AI Agent Liability
The report identifies that no current law establishes responsibility when an AI agent makes autonomous purchases. Chargeback and dispute rules were designed for human consumers shopping at human speeds, not software executing thousands of payments per hour. The boundary between card and crypto systems is fading as card initiatives including the Trusted Agent Protocol, Agent Payments Protocol and Visa Intelligent Commerce add stablecoin support while crypto projects adopt trust and identity verification processes long used by card networks.
Visa Joins Open USD Consortium and Reports $7 Billion Stablecoin Settlement
At June's Payments Forum, Visa connected its Intelligent Commerce platform to an OpenAI partnership enabling ChatGPT agents to purchase items. The company reported approximately $7 billion in annual stablecoin settlement volume. Days after the announcement, Visa joined Mastercard, Coinbase and over 140 firms in the Open Standard consortium supporting Open USD. The payment shift extends beyond card companies, with Binance directing its next growth initiative toward payments over trading and Interactive Brokers adding two-way stablecoin transfers this month.
FAQ
What payment model did Visa and Artemis propose in their joint report this week?
Visa and Artemis proposed a hybrid model where card networks handle consumer purchases while stablecoins process small, frequent machine-to-machine transfers under one dollar between software programs.
How much transaction volume has the x402 protocol processed since May 2025?
The x402 protocol processed approximately $15 million across 109.6 million payments since May 2025, primarily on Base, Solana and Polygon networks, with most transfers sitting well under one dollar.
What regulatory gap did the Visa-Artemis report identify for AI agent payments?
The report identified that no current law establishes who is responsible when an AI agent makes autonomous purchases, and existing chargeback and dispute rules were designed for human consumers rather than software executing thousands of payments per hour.
