Nearly nine in 10 qualifying wallets first funded in Q2 2026 got another deposit within 30 days. But a study of Base settlements found much of the activity involved connected wallets.
On Base, researchers examined 136.7 million x402 settlements worth $44.1 million. They estimated that potentially genuine economic activity could have reached $20.26 million. Another 21.2% of transactions were fictitious, including self-payments and closed loops. The figures put a commercial caveat on the protocol's growing payment activity: transaction counts alone do not establish independent customer demand.
Chainalysis data show developers are funding x402 wallets more often. Of wallets active on at least two separate days, 89.3% of those first funded in the second quarter of 2026 received another deposit within 30 days. For the same qualifying group first funded in the third quarter of 2025, the share was 70.6%.
Payments get faster
Newly funded wallets are also making their first payment sooner. Two-thirds did so within an hour in the third quarter, up from 4% a year earlier, according to Chainalysis. The median time from initial funding to first payment fell to 18 minutes from 73 days in late 2025.
These measures track wallet activity, not unique people or businesses. Chainalysis attributed the repeat deposits to more regular spending by developers. The firm also found that the share of agent wallets making a first payment within an hour of funding rose from 4% in the third quarter of 2025 to 67% in the third quarter of 2026.
Coinbase published x402 in May 2025 as an on-chain payment protocol for AI agents and other autonomous software. It revives HTTP's 402 "Payment Required" status: a service requests payment, software settles it in cryptocurrency, and the client gets access to the requested resource. An agent can pay for an API call or a model response without a person approving each transaction. It can also pay for data or computing capacity.
Cumulative x402 transactions on Base rose from 58,800 in the third quarter of 2025 to 138.5 million in the third quarter of 2026, Chainalysis reported. In a separate 30-day measure for July, the protocol handled about 75 million payments worth roughly $24 million. More than 87% of payments analyzed over the 16 months through September 2026 were between $0.0001 and $0.10. One- to 10-cent payments were the most common, according to Chainalysis' report.
Small payments need customers
At those amounts, the business case hinges on whether a service can collect frequent payments at a cost it can support. Conventional card networks can make tiny charges uneconomical because of fixed processing fees and operating costs. Stablecoin settlement can handle smaller payments, but low transaction values leave little room for computing and infrastructure costs.
x402 payments cover AI model inference and code reviews. They also go toward financial intelligence and data retrieval. Inference is when one application pays another AI system to generate a response. Chainalysis found that inference had become one of x402's largest transaction categories. Subsequent payments also flowed to information services and AI infrastructure.
That creates a potential way for developers to earn per-request revenue instead of relying on subscriptions or enterprise contracts. But the activity does not show, by itself, that services have paying customers independent of their operators. A separate liquidity analysis drew a related distinction: an aggregate figure does not establish what an individual user can access.
Geography and proof
Chainalysis estimates that U.S.-attributed wallets account for about 30% of geographically identified transactions between AI agents and service providers. Their share is higher for inference: American-linked wallets generated 88.1% of those identified payments. U.S. wallets also represented 41.8% of payments to agent frameworks and development tools. These figures cover identifiable x402 transactions and geographically attributed wallets, not the entire global market for AI-agent payments.
Chinese-linked wallets followed a different pattern. They accounted for about 90% of geographically attributed payments for security and code-analysis services, much of it directed to AgentLISA, which audits blockchain smart contracts using AI. Their share of identified inference transactions was 0.1%. Chainalysis suggested that greater reliance on open-weight models such as DeepSeek and Qwen could help explain the gap. Those models can run without an external provider charging for every request.
The academic study puts another limit on headline transaction totals. It classified 63.78% of settlements as occurring within wallet clusters connected through funding relationships. That category can include related operators as well as legitimate commerce. Researchers identified about $187,861 in payments reaching recognizable services without detected buyer-seller connections. They also found that 25,163 listed Base resources mapped to 811 distinct payment recipients. Of those recipients, 249 earned at least $10 during the observation period.
For x402 businesses, recurring receipts from independent customers would provide stronger evidence of commercial traction than wallet refills or raw settlement counts. CoinDesk reported that more than 87% of payments analyzed by Chainalysis were worth between one-hundredth of a cent and 10 cents. The protocol uses web requests to initiate payment before delivering a resource. The actual charge can be a fraction of a cent, making the cost of serving each request and the source of each payment central to whether the model can support a business.
