A striking 76% of AI agent transactions fall below the $0.30 fee charged by traditional card networks, underscoring the economic imperative for crypto rails in this emerging market. A report from Keyrock, released in May 2026, documented that AI agents have facilitated more than $73 million across approximately 176 million transactions since May 2025, with an impressive 98.6% of these payments made in USDC.
The report emphasizes the rapid evolution of machine-to-machine payment systems, which have moved from theoretical concepts to a functioning ecosystem within a year. Major players like Coinbase and Stripe have strategically positioned themselves across five of six key layers in the payment stack. This layered approach involves protocols that manage authorization, settlement, and governance, creating a complex environment where capturing value is crucial.
Keyrock’s analysis reveals that the average transaction size has stabilized around $0.48, with Coinbase’s x402 protocol repurposing the long-dormant HTTP 402 status code to facilitate stablecoin transactions. Additionally, Stripe and Tempo’s Machine Payments Protocol (MPP) has emerged as a versatile solution, accommodating various payment methods, including stablecoins and credit cards.
The dominance of USDC in this sector raises significant concerns about systemic risk. With 98.6% of payments executed in this stablecoin, any regulatory challenge or technical failure affecting Circle, the issuer of USDC, could threaten the entire agent payments ecosystem. This scenario highlights the need for diversification in payment methods, a feature that currently seems lacking.
Incumbents in the financial sector have been active in solidifying their positions through substantial acquisitions, with over $8 billion invested in various deals over the past year. For example, Capital One's acquisition of Brex for $5.15 billion and Mastercard’s purchase of BVNK for $1.8 billion illustrate a growing interest in deeper integration within this evolving market. The urgency to build out capabilities reflects a recognition that the future of commerce may increasingly rely on automated transactions facilitated by AI agents.
Despite the promising growth of AI agent transactions, regulatory uncertainty remains a significant concern. Key European regulations, including the MiCA, the GENIUS Act, and the EU AI Act, will take effect throughout 2026, yet none of these frameworks adequately address machine-to-machine payment transactions. This regulatory gap leaves unresolved liability issues; merchants currently bear the risk of chargebacks with credit card transactions, while stablecoin transactions, once completed, are irreversible.
The report notes that while AI agents dominate on-chain activity—accounting for a significant portion of transactions on networks like Gnosis and Base—the current use cases are largely extractive, focusing on arbitrage and volume farming. The anticipated shift towards productive agent commerce, where AI agents pay for services that provide real value to end users, remains a work in progress.
The failure of OpenAI's ChatGPT Instant Checkout, which was discontinued earlier this year, further illustrates the complexities of integrating AI into transactional frameworks. The product struggled to gain traction, lacking essential features such as sales tax collection and multi-item cart support, reinforcing the idea that agents are more effective when operating through protocol endpoints rather than traditional e-commerce interfaces.
As the machine economy develops, the infrastructure necessary for meaningful commerce is in place, but the regulatory clarity needed to support its growth is absent. Until these fundamental issues are resolved, the future of AI agent transactions remains uncertain, even as the volume of activity continues to rise.
The stories that move AI & crypto markets — before the market reacts.
Free. 7am ET. Five stories. 62,400 readers.



