A new report from Keyrock estimates AI agents settled over $73 million on blockchains in the last year, signaling a foundational shift in machine-to-machine payments.
A new report from Keyrock estimates AI agents settled over $73 million on blockchains in the last year, signaling a foundational shift in machine-to-machine payments.

AI agents settled more than $73 million across 176 million blockchain transactions over the past year, with 98.6% of the volume flowing through Circle’s USDC stablecoin. The activity highlights a growing race by firms including Coinbase, Google, and Visa to build payment infrastructure for a machine economy that traditional card networks are ill-equipped to handle.
“Card networks and bank transfers can't handle instant, programmable, machine-to-machine value transfer. Stablecoins on blockchain can,” Idan Ofrat, co-founder and chief product officer at Fireblocks, said in a statement. “Digital assets aren't just a good fit for agentic payments. They're the only fit.”
The economics explain the trend, with 76% of agent transactions falling below the 30-cent fee floor common in card payments, according to a report from crypto investment firm Keyrock. Most automated payments range from one to 10 cents. Coinbase’s x402 protocol has emerged as a key rail, processing over $50 million in USDC payments on the Base network to date.
The development creates a significant new use case for stablecoins as functional, high-velocity money rather than just a trading instrument. With Gartner projecting AI agents could handle $15 trillion in purchases by 2028, the battle for the underlying payment rails is intensifying, though regulatory frameworks like MiCA and the GENIUS Act, expected in mid-2026, do not yet directly address autonomous machine payments.
The race to build the plumbing for the machine-to-machine economy has drawn in some of the largest players in tech and finance. Coinbase’s x402 protocol, which leverages the dormant HTTP 402 "Payment Required" status code, allows agents to pay for services like API calls directly with USDC on Base, bypassing traditional accounts. The protocol has seen accelerating adoption, processing over $50 million in payments across more than 2,000 integrated APIs.
Competitors are moving quickly. Stripe launched its Machine Payments Protocol (MPP) on its proprietary Tempo blockchain, while Google introduced a system called AP2 focused on delegated spending authorization. Even payment giant Visa has extended its card network with tokenized credentials designed for AI-driven commerce. Institutional players are also joining, with digital assets platform Fireblocks launching a suite for x402 and joining the protocol's foundation alongside Google, AWS, and Mastercard.
The infrastructure is extending to the application layer. Trust Wallet, one of the most widely used self-custody wallets, integrated Binance's fork of the x402 protocol to allow AI agents to make payments directly from user wallets without exposing private keys. This self-custody approach reduces counterparty risk compared to solutions that require funds to be held by a third-party platform.
The impact is most visible with major AI service providers. OpenRouter, an aggregator that routes requests across dozens of AI models and handles roughly $1 billion in inference volume annually, is transitioning its settlement operations to the x402 protocol. This move could significantly increase transactional demand for USDC, cementing its role as the functional currency for AI agents. However, the heavy reliance on a single stablecoin also introduces concentration risk, creating a dependency on a single issuer, Circle.
This article is for informational purposes only and does not constitute investment advice.