U.S. Leads AI Model Spending as AI Agents Execute 138 Million Crypto Payments

Artificial intelligence agents have been involved in 138 million cryptocurrency payments, highlighting the growing use of digital assets for transactions linked to AI services and model access, according to new research from Chainalysis.

The activity is heavily concentrated in wallets connected to the United States. U.S.-linked wallets represented slightly more than 88% of tracked payments to AI models, while wallets associated with China accounted for just 0.1%, the research found.

The gap points to a pronounced geographic imbalance in the emerging market for machine-to-machine payments. Although AI systems can be developed and accessed globally, the crypto activity identified by Chainalysis is overwhelmingly tied to the United States. The findings do not necessarily measure total AI use or all payments made for artificial-intelligence services. Instead, they reflect transactions that the research was able to track through cryptocurrency wallets and associate with AI models.

That distinction is important because payments connected to AI can take several forms. A user may pay for access to an AI model, an automated application may purchase data or computing resources, or an agent may execute a transaction on behalf of a person or business. Blockchain-based payments are often presented as a possible way for software agents to transact without relying on conventional bank accounts, card networks or manual approval for every payment.

The Chainalysis figures offer an early indication of where that activity is developing most visibly. The United States’ more than 88% share places it far ahead of every other market identified in the research. China’s 0.1% share, by comparison, suggests that only a very small proportion of the tracked crypto payments linked to AI models came from China-associated wallets.

The numbers should not be read as a direct ranking of national AI capabilities. Wallet attribution is not always equivalent to the physical location of a user, company or developer. A wallet may be controlled by a global business, operated through a service in another jurisdiction, or linked to users whose activities cross borders. Likewise, the research captures identifiable blockchain transactions rather than every payment made through traditional financial channels or private systems.

Even with those limitations, the scale of the reported activity is notable. One hundred and thirty-eight million payments indicates that AI-related crypto transactions are not confined to isolated demonstrations or small-scale experiments. It suggests a substantial volume of activity involving automated or semi-automated payments, though the available findings do not specify the total dollar value, average transaction size or period covered by the research.

The data also underline the importance of infrastructure in the development of AI agents. For an automated system to operate independently, it may need more than the ability to generate text, analyze information or complete digital tasks. It may also need a way to identify a service, authorize a payment and receive the corresponding resource. Cryptocurrency networks can provide a programmable payment layer for those functions, particularly where transactions are conducted online and across borders.

At the same time, crypto payments introduce challenges that may limit broader adoption. Transactions can be difficult to reverse, wallet ownership may be unclear, and automated systems must be protected against unauthorized transfers, fraudulent instructions and software errors. Businesses using AI agents would also need to address compliance requirements, financial controls and the risk that a system could make payments beyond its intended authority.

The geographic concentration in the Chainalysis research may reflect several factors, including the location of companies building AI products, the availability of crypto infrastructure and the willingness of businesses and users to experiment with automated payments. It may also be influenced by how blockchain activity is attributed and which services are visible to researchers. The report’s figures therefore provide a picture of tracked on-chain activity rather than a complete census of the global AI economy.

For the crypto industry, the findings reinforce the idea that AI agents could become an important source of transaction demand. For AI companies, they raise questions about how automated systems will obtain services and interact with commercial platforms as their capabilities expand. The two sectors are increasingly being discussed together because both depend on software that can operate at high speed and, in some cases, with limited human intervention.

The immediate picture from Chainalysis is nevertheless clear: crypto payments associated with AI models have reached a substantial volume, and the activity recorded by the research is dominated by U.S.-linked wallets. Whether that pattern persists will depend on regulation, the growth of AI services, the reliability of automated payment systems and the extent to which businesses adopt crypto as a practical settlement method for software-driven transactions.