In Chainalysis’s 2026 North America study, over 138 million x402 payments were recorded on Base. These transactions represent a significant volume but are typically small in value, with U.S.-linked wallets making up the majority of geolocated transactions for model inference.
The report tracks various categories of recipients, including agent frameworks, tool catalogues, and payment rails. However, calling all these transactions 'AI productivity spending' would be misleading, as some payments might involve speculative activities or recurring software workflows.
Volume and Typical Payment Size
Chainalysis extended its examination to September, providing a broader view of the payment sizes. Applying the 87% micropayment share from this extended sample to all transactions would be inaccurate without matching datasets and definitions across periods.
The study aims to test the persistence of improvement by tracking wallets that return and refill. If these wallets continue to fund services affiliated with their operators, it suggests operational reuse rather than a broad market for AI productivity spending.
Attribution challenges arise as wallets might be labeled agents based on protocol usage, but the actual payer could be a human script or promotional service. Chainalysis measures observable interactions, not private decision processes behind transactions.
Small payments can accumulate to significant costs through frequency. For instance, sending a five-cent request once per second would result in an hourly spend of $180, highlighting the need for rate limits and spending caps on AI agents.
Payment Recipients and Wallet Behavior
The success of agent payments depends on reducing onboarding friction, reaching a useful market, and settling at acceptable total costs. The 138 million count indicates that software is using the payment rail, but future data will show its economic impact.
The report also covers payments to game-related services, token issuers, and gambling services, adding complexity to categorizing these transactions as AI productivity spending. These categories underscore the need for careful analysis of payment protocols and recipient activities.