Artificial intelligence (AI) agents are currently moving past simple peer-to-peer conversation to settle bills for services, acquire datasets, and purchase access to AI model tokens. Over 138 million transactions have been logged via the x402 protocol running on Coinbase’s Base network, highlighted in a report by Chainalysis titled Inside North America’s $3 trillion crypto economy: Agentic commerce accelerates.
This emerging ecosystem is heavily dominated by the United States. Wallets tied to the U.S. represented 88.1% of all geographically traceable transactions for AI model inference, contrasted with a mere 0.1% originating from Chinese-connected addresses, according to data from the blockchain research firm.
Eric Jardine, research director at Chainalysis, explained to CoinDesk that a plausible theory is that developers in China lean more heavily toward open-weight models like Qwen and DeepSeek, lowering the requirement to buy metered inference solutions.
These metrics, published on Wednesday, arrive as major payment providers including Mastercard, Visa, and Stripe have thrown their weight behind the x402 standard, alongside Coinbase opening pathways for business clients to accept funds directly from automated agents.
Animoca Brands co-founder and chairman Yat Siu recently shared with CoinDesk his projection that the next few years could see up to 100 billion autonomous AI agents executing web-based transactions using cryptocurrency. Furthermore, stablecoins may serve as the onboarding vehicle for crypto’s next billion participants. Stephanie Cohen, Cloudflare’s chief strategy officer, stated to CoinDesk that stablecoins fit the exact profile of transactions that AI agents will initiate initially.
In contrast to standard card infrastructure, stablecoins support sub-penny accounting, enabling software agents to purchase computing power, data feeds, and supplementary utilities free from human oversight. The recent Chainalysis data points to increased utilization of these financial rails, although raw transaction volume alone does not confirm authentic commercial demand.
A significant portion of this activity moves through x402, an incubator payment framework created via Coinbase that enables AI programs to autonomously settle digital service fees using crypto assets.
Chainalysis revealed that more than 87% of tracked x402 payments fell between one-hundredth of a cent and 10 cents, rendering them far too minuscule for standard credit card processors to handle profitably.
Absolute values continue to stay modest. Back in July, x402 handled roughly 75 million transactions valued at an aggregate of $24 million across a 30-day window. Prior studies also voiced skepticism regarding artificial traffic observed on the network.
Among digital wallets initialized during the second quarter of 2026, 89.3% secured fresh top-ups within a 30-day window, compared to 70.6% for counterparts established in Q3 2025. This evaluation left out wallets that only executed transfers on a single calendar day.
Additionally, the timeframe separating a wallet’s funding event from its inaugural x40 payment dropped drastically, plunging from a 73-day average in late 2025 down to just 18 minutes by Q3 2026.
Meanwhile, activity originating from China displayed a distinct trend. Accounts linked to China drove nearly 90% of localized payments toward code analysis and security utilities, with a large share routed to AgentLisa, an AI-driven smart contract auditing network.
Jardine cautioned observers against treating that specific disparity as definitive proof that open-source AI is hurting the Asian payments ecosystem.
He noted that while it remains a viable hypothesis, they would not go so far as to claim open-source software is starving the on-chain inference sector in Asia.
The reported metrics track transparent x402 transfers rather than total worldwide spending on AI infrastructure.
Tejwani noted in the study that AI agents are evolving into genuine economic participants, where every single agent acts functionally as a brand-new blockchain user.
While monetary amounts remain low today, the metrics compiled by Chainalysis indicate that automated algorithms are forging their own transaction habits, with the United States holding the premier spot in paid AI services.
Originally published at https://www.coindesk.com/business/2026/10/14/ai-agents-make-138-million-crypto-payments-as-u-s-dominates-spending-on-ai-models.
