AI Agents Process 138M Crypto Payments, Most Tiny

AI agents made 138 million crypto payments in a year — and most were for pennies

Crypto payments made by autonomous AI software are starting to show up at scale on-chain, even if the dollar amounts remain small.

Activity routed through x402, a payments protocol developed by Coinbase, surpassed 138 million transactions on the Base blockchain by July 2026, according to data cited from blockchain analytics firm Chainalysis. The count represents a sharp rise from roughly 1,500 transactions in mid-2025.

The vast majority of these payments were tiny. Chainalysis found that more than 87% of x402 transactions landed in the $0.01 to $0.10 range, indicating that early “agent commerce” is less about consumer shopping and more about high-frequency machine-to-machine settlement.

These transactions are typically payments for digital services that software agents consume continuously—such as API access, data, compute, and AI inference. In many cases, both sides of the transaction are programs rather than humans, a pattern Chainalysis described as AI agents becoming “economic actors,” with each agent effectively representing a new “user” on a blockchain network.

Even with the transaction volume, the economic footprint is still limited. x402 moved about $24 million over 30 days in July, a level that highlights the gap between large on-chain activity counts and relatively modest payment value.

Stablecoins appear to be the main settlement rail for this type of payment behavior. Circle’s disclosures show USDC accounted for 98.8% of agent-driven transaction volume, and Circle CEO Jeremy Allaire previously said the share reached 99.3% for AI-agent payments using x402 in the second quarter of 2026.

Chainalysis also broke down where spending is coming from. Wallets linked to the United States accounted for 88.1% of AI inference service spending tracked in the dataset, underscoring the U.S. lead in the market for paid AI services.

Still, Chainalysis cautioned that raw transaction counts can be misleading. It noted that a late-2025 surge in activity was tied in part to meme coin activity, complicating efforts to separate experimental or opportunistic on-chain behavior from sustained demand for real AI services.

Taken together, the numbers point to a practical near-term use case for crypto: micropayments between software systems that need a standard way to pay for online resources in small increments. The challenge ahead is measuring how much of this growth reflects durable utility versus transient on-chain noise.

Source: Chainalysis (as reported by CoinDesk)