A new report from crypto trading and investment firm Keyrock reveals that artificial intelligence agents are increasingly relying on blockchain rails for autonomous payments, with settlement volumes reaching $73 million across 176 million transactions between May 2025 and April 2026. While the dollar amount is tiny compared to traditional finance—Visa processes $14.5 trillion annually—the rapid formation of infrastructure by global tech and payments giants signals a transformation in machine-to-machine commerce.
Major players race to build infrastructure. Coinbase has introduced the x402 protocol, a crypto-native system enabling AI agents to pay directly with USDC for services like blockchain analytics without accounts. Stripe launched the Machine Payments Protocol (MPP) on its Tempo blockchain, Google rolled out AP2 for delegated spending authorization, and Visa extended its card network with tokenized credentials for AI-driven commerce. These moves indicate the market is leaving its experimental phase behind.
Economics favor stablecoins. According to the report, 76% of agent transactions fall below the $0.30 fixed-fee floor typical of card payments, with most ranging between one and ten cents. Stablecoin settlement on blockchains like Base and Tempo costs fractions of a cent, making traditional rails impractical. As a result, 98.6% of machine payments now settle in USDC, the stablecoin issued by Circle—a dominance that brings both efficiency and concentration risk.
Ambitious forecasts point to explosive growth. Gartner projects AI agents could intermediate $15 trillion in purchases by 2028, while McKinsey estimates retail agentic commerce may reach $3–5 trillion by 2030. Such growth rates would outpace even stablecoins’ breakout years. However, the report cautions that regulatory frameworks like Europe’s MiCA, the U.S. GENIUS Act, and the EU AI Act—all expected around mid-2026—currently fail to address autonomous machine-to-machine transactions, potentially creating legal uncertainties around liability and agent identity.
