Autonomous AI agents have executed over 138 million cryptocurrency payments within a year using the x402 protocol, a Coinbase-incubated system designed for automated software transactions. This activity is primarily concentrated on Base, the layer-2 network developed by Coinbase, where cumulative payments surged from approximately 1,500 in mid-2025 to over 138 million by the third quarter of 2026.
The vast majority of these transactions are micro-payments, with over 87% of them falling within the range of 0.01 cents to 10 cents. These small-scale transfers are utilized by AI agents to acquire computing power, real-time web data, and various coding tools. According to Chainalysis, such micro-transactions are currently supported by stablecoins, whereas traditional card payment networks struggle to process them.
Geographically, the United States leads this sector, with American-linked wallets responsible for 88.1% of identifiable payments related to AI model inference. In contrast, Chinese-linked wallets account for only 0.1% of inference payments, though they represent the majority of spending on code analysis and security services. Chainalysis notes that these figures reflect only identifiable x402 activity.
Major financial and technology firms are actively building infrastructure to support these autonomous transactions. Visa, Mastercard, Stripe, and Google have introduced initiatives for agent payments, while Coinbase has launched tools for agents to transact and for businesses to accept such payments. BlackRock is also monitoring the sector, identifying AI agents as a potential source of demand for frequent, low-value transactions.
Despite the rapid growth, analysts advise caution regarding the interpretation of these metrics. Research from TRM suggests that a significant portion of the observed activity might stem from simple scripts or self-dealing rather than fully autonomous agents. This distinction highlights the difficulty in quantifying the actual scale of agentic commerce based solely on raw transaction counts.
