AI Agents Make Millions of Micropayments Through x402
Chainalysis already sees 138 million transactions for AI services on Base through x402. Small stablecoin payments in particular seem well suited for machine-to-machine use.
Key Takeaways
- AI agents use the x402 protocol to pay crypto for small payments to data, compute, and security services.
- Chainalysis reports more than 138 million transactions on Coinbase’s Base blockchain, with over 87% of them between 0.01 cent and 10 cents.
- The United States dominates this market, while overall activity is still small and may be partly artificial.
AI agents are no longer just chatting or generating. They also pay for data, compute, and security services. According to Chainalysis, more than 138 million transactions have already been processed through the x402 protocol on Coinbase’s Base blockchain. That shows autonomous software is increasingly using crypto for small payments that are hard to handle through traditional payment networks.
Micropayments for AI Services
The core of the system is simple: an AI agent can pay on its own to access a digital service. In practice, this often means amounts ranging from fractions of a cent to a few cents. Chainalysis found that more than 87% of x402 payments were between one-hundredth of a cent and 10 cents.
That fits the kind of services AI agents use. They buy access to models, data, and other digital services, without a human needing to step in every time. According to the researchers, that kind of small-scale usage is exactly why crypto and stablecoins are a good fit for this market. Coinbase’s agent accounts also show how quickly this payment infrastructure is developing for software that trades and pays on its own.
The US Is Leading
Chainalysis says the United States clearly dominates this segment. Of the geographically identified payments for AI model inference, 88.1% came from wallets with a US link, while wallets with a China link made up just 0.1%. Eric Jardine, head of research at Chainalysis, said one possible explanation is that Chinese developers more often use open-weight models like DeepSeek and Qwen, which means they need paid inference services less often.
For now, activity remains small in absolute terms. In July, x402 processed about 75 million payments worth a total of $24 million (€21.4 million) over 30 days, according to the research. At the same time, signs emerged that part of the activity may be artificial, although the figures only show visible transactions on x402 and not the full global market for AI payments.
Why This Matters
For European crypto readers, this is especially interesting because it shows where stablecoins and crypto rails are actually being used: not just for trading, but also for machine-to-machine payments. That could matter for payment companies, developers, and firms looking at new forms of digital services. At the same time, the market is still small, and it is not clear how much of the activity reflects real commercial demand.
