AI agents settle $73M on crypto rails, Keyrock report shows
In brief
- AI agents settled $73 million across 176 million blockchain transactions in 12 months
- 76% of machine payments fall below 30-cent card network fees, making crypto rails economical
- USDC dominates machine payments at 98.6% market share, creating concentration risk
The economics of machine payments
The gap between blockchain settlement and card rails comes down to transaction size. Some 76% of agent transactions fall below the 30 cent fixed-fee floor common in card payments. Most payments ranged between one and 10 cents, making traditional rails impractical for automated software agents buying services or data.
Stablecoin settlement on blockchains like Base and Tempo costs fractions of a cent. That math is unbeatable for high-frequency, low-value transactions.
"Crypto rails and stablecoins are emerging as the preferred settlement layer, and the economics help explain why." — Keyrock report
The infrastructure race
Global firms such as Coinbase (COIN), Stripe, Google (GOOG) and Visa (V) all rolled out competing systems for machine-to-machine payments. Coinbase's x402 protocol allows AI agents to pay directly with USDC for services such as blockchain analytics or cloud infrastructure without creating accounts or subscriptions. Stripe, with its Tempo blockchain, launched a competing framework called Machine Payments Protocol (MPP).
Google introduced AP2, a system focused on delegated spending authorization for AI agents. Visa has extended its card network with tokenized credentials designed for AI-driven commerce.
The race reflects the scale at stake. Gartner projects AI agents could intermediate $15 trillion in purchases by 2028. McKinsey estimated retail agentic commerce could reach $3 trillion-$5 trillion by 2030.
The USDC concentration problem
Currently, 98.6% of machine payments settle in USDC, the stablecoin issued by Circle (CRCL). That solidifies Circle's position in crypto payments, but also introduces risk of concentration, creating dependency on a single issuer.
Regulatory clarity remains absent. MiCA in Europe, the U.S. GENIUS Act and the EU AI Act are all expected to take effect around mid-2026, yet none of them directly address autonomous machine-to-machine transactions or questions around liability and agent identity.