Chainalysis finds AI agents executed more than 138 million x402 payments on Base, mostly tiny purchases

Chainalysis’s 2026 North America study found that more than 138 million x402 payments moved across Coinbase’s Base network over its reporting window, and that the vast majority were tiny transactions directed at services tied to AI.

The firm’s analysis places most payments in micro-value bands and breaks down recipients into categories that include AI model inference, code analysis, search and data feeds, agent tools and payments infrastructure.

Chainalysis reported that over 87% of payments in its extended sample were between $0.0001 and $0.10, and that US-linked wallets accounted for 88.1% of geolocated payments labelled as AI model inference.

The study also documented rapid wallet behaviour: qualifying wallet cohorts refilled within 30 days at rates the firm said rose from 70.6% to 89.3%, while the median time from funding to a first payment shrank from 73 days to 18 minutes.

Chainalysis highlighted a surge in near-immediate spending, noting the share of payments occurring within one hour of funding climbed sharply between Q3 2025 and Q3 2026. The report situates the enormous transaction count in operational terms rather than pure retail commerce.

Chainalysis identified categories of recipients that suggest an economy of machine-to-machine payments: AI model inference calls, automated code-analysis tools, search and data feed services, agent toolkits and payments routing.

That pattern, the firm argues, indicates many wallets were executing tiny payments to buy modular compute or data services used by automated agents rather than funding traditional consumer purchases. Chainalysis also traced the origin of early activity on Base to experiments and memecoin-driven minting.

The firm named a pay-to-mint experiment called PING and other memecoin demand as drivers of initial transaction volumes that later abated, leaving behind smaller-value payments linked to AI services. Chainalysis said those early speculative mint events generated transaction counts that did not necessarily reflect a durable service economy.

The study’s timing and sample windows produce slightly different numeric framings across outlets: Chainalysis described the 138 million figure as cumulative to the end of its 12-month study period ending 1 July 2026 in one account, while other summaries present the total as the firm’s count “by the end of its 2026 reporting period.” The report also compared cohorts across quarters, showing a movement toward faster funding-to-spend behaviour with the newest cohorts in Q2 2026 exhibiting much shorter funding-to-first-payment intervals.

Independent outlets that republished or summarized Chainalysis’s findings emphasised different elements: some foregrounded the headline transaction count and the link to AI agents, others highlighted the tiny size of most payments or the memecoin-driven origins.

All sources in this pack agree on the scale of the x402 activity and on the prominence of micro-payments; differences lie mainly in which metrics each outlet emphasised.

Chainalysis’s breakdown of recipient categories and wallet behaviours presents a picture in which automated systems and agent tooling appear to be a primary consumer of micropayments on Base, supported predominantly by US-linked wallets.