TL;DR
Chainalysis recorded more than 138 million transactions carried out by artificial intelligence agents through Coinbase’s x402 protocol.
U.S.-linked wallets accounted for 88.1% of geographically identified payments for access to AI models, compared with China’s 0.1% share.
More than 87% of the payments analyzed were worth less than $0.10, while the average time between a wallet receiving funds and making its first transaction fell to 18 minutes.
Chainalysis identified more than 138 million transactions processed through the x402 protocol , which artificial intelligence agents use to pay for digital services with cryptocurrencies.
According to a report from the blockchain analytics firm, these automated systems are using digital assets to purchase data, computing power, and access to AI models without requiring human intervention for each transaction.
The United States accounts for 88.1% of geographically identified payments for AI model inference services. China, by comparison, represents just 0.1% of that activity.
The findings cover transactions tracked through x402 rather than the entire global market for payments made by AI agents.
Chainalysis Analyzes 138 Million Transactions Through x402
The x402 protocol, initially developed by Coinbase, allows applications and AI agents to pay automatically for digital services using cryptocurrencies.
One of its main features is the ability to process micropayments that would be economically impractical through traditional card networks.
According to Chainalysis, more than 87% of the payments analyzed ranged from $0.0001 to $0.10 .
In July, x402 processed approximately 75 million payments worth a total of $24 million over a 30-day period.
The difference between the number of transactions and the total amount transferred is due to the small size of individual payments .
The report also identified changes in the behavior of wallets using the protocol.
Among addresses first funded during the second quarter of 2026, 89.3% received additional funding within the following 30 days . That figure stood at 70.6% for wallets initially funded in the third quarter of 2025.
The comparison excluded addresses that made payments on only one day.
The average time between a wallet receiving funds and making its first x402 payment also dropped from 73 days in late 2025 to just 18 minutes in the third quarter of 2026 .
United States Dominates Payments for AI Models
Chainalysis also found differences in the types of services purchased by agents linked to different regions.
U.S.-linked wallets accounted for 88.1% of identified payments for artificial intelligence models , while addresses associated with China represented just 0.1%.
Eric Jardine , Head of Research at Chainalysis, suggested that Chinese developers may be more likely to use open-weight models such as DeepSeek and Qwen , reducing their need to pay for inference services.
However, he cautioned that the data does not establish that the availability of open models is undermining Asia’s blockchain payments market.
China shows a different pattern in other services. Wallets linked to the country account for approximately 90% of geographically identified payments for security and code analysis .
A substantial share of those transactions goes to AgentLisa , an AI-powered smart contract auditing platform.
The development of automated payments has also attracted companies such as Visa, Mastercard, and Stripe , which have backed x402. Coinbase, meanwhile, has started allowing its business customers to accept payments from AI agents.
Although transaction volumes are growing, Chainalysis warns that transaction counts alone cannot establish how much activity represents genuine commercial demand . Previous research has also raised questions about artificial activity on the protocol.
