AI Agents Made 138 Million Crypto Payments in a Year, Chainalysis Finds
Key Takeaways
- •x402 payments on Base increased from roughly 1,500 in mid-2025 to over 138 million by the third quarter of 2026.
- •More than 87% of the transactions were worth between $0.0001 and 10 cents, a range suited to stablecoin payments.
- •U.S.-linked wallets represented 88.1% of geographically identifiable payments for AI model inference, compared with 0.1% for China.
- •Visa, Mastercard, Stripe, Coinbase and Google have introduced or supported infrastructure for payments made by AI agents.
- •TRM warned that transaction totals may include automated scripts or self-dealing, potentially overstating genuine agentic commerce.

AI agents are beginning to spend real money, and increasingly they are paying in crypto.
That is the picture from a new Chainalysis report, which found that autonomous AI agents have made more than 138 million cryptocurrency payments through x402, the Coinbase-incubated protocol — named for HTTP status code 402, the web's long-unused "Payment Required" response — that lets software pay for online services automatically.
The report, part of the blockchain analytics firm's annual Geography of Crypto series and focused this year on North America's $3 trillion crypto economy, frames "agentic commerce" — payments initiated and settled by software rather than people — as one of the sector's fastest-growing use cases.
The growth has been steep. Cumulative payments through x402 on Base, Coinbase's Ethereum layer-2 network, rose from roughly 1,500 in mid-2025 to more than 138 million by the third quarter of 2026, as developers increasingly hand AI agents their own budgets to buy computing power, live web data and coding tools on the fly.
The payments are tiny: more than 87% fell between a hundredth of a cent ($0.0001) and 10 cents — the kind of micro-transactions that stablecoins can handle but traditional card rails cannot, since card payments were built around human account holders and per-transaction costs that make sub-cent purchases uneconomical.
The United States dominates the activity. American-linked wallets accounted for 88.1% of geographically identifiable payments for AI model inference — the computing task of running AI models — Chainalysis found, compared with just 0.1% for China. Chinese-linked wallets, however, made up the bulk of spending on security and code-analysis services.
The firm cautioned against reading too much into the gap, noting that its figures capture only identifiable x402 activity rather than the entire global market.
The findings land as the infrastructure for agent payments rapidly matures across both crypto and traditional finance — a sign of how far payment providers are moving to accommodate customers that cannot open bank accounts or click checkout buttons of their own. Payments giants Visa, Mastercard and Stripe have all moved to support AI-agent payments. Coinbase rolled out a tool letting agents trade and transact and has begun letting business customers accept agent payments, while Google unveiled a dedicated agent-payments protocol of its own.
BlackRock, the world's largest asset manager, has separately flagged autonomous agents as a potential next wave of crypto demand. Will Su, the firm's head of digital assets research, told Chainalysis in the report that the company is watching the space closely, "with AI agents potentially creating meaningful demand for frequent, low-value payments around the clock."
A note of caution hangs over the numbers, however. Earlier research from TRM, a rival crypto tracing and blockchain analytics company, suggested much of the apparent agent activity on these rails may be scripts or self-dealing rather than genuine autonomous agents — a reminder that raw transaction counts can overstate how much real agentic commerce is actually taking place. How follow-up research reconciles that gap is likely to shape how much weight headline volumes deserve as the sector grows.
Still, the trajectory points one way, and the thesis increasingly floated across the industry is that crypto's next wave of users may not be people at all.