Coinbase has unveiled AiFi, a new initiative designed to address the financial needs of AI agents. The announcement argues that autonomous agents require robust, always-on financial infrastructure to transact effectively, and AiFi is positioned as a framework for programmable financial systems tailored to machine-driven economic activity. The move signals a broader shift toward integrating AI into economic systems and could reshape how value is transferred and managed.
Separately, Coinbase reported that AI agents have executed more than 205 million transactions through the x402 protocol, with cumulative volume reaching $53 million. Coinbase’s network accounted for the largest share of that activity at 67%, underscoring its early lead in a niche that is still taking shape.
The x402 protocol is an open standard that enables AI agents and automated applications to initiate payments directly over HTTP without requiring a human in the loop. It effectively turns a standard web request into a payment instruction, allowing machines to pay for services, data, or computational resources on demand. Because it leverages existing web infrastructure, it lowers the barrier for developers integrating payments into AI workflows.
Coinbase’s dominance in transaction count reflects its early integration of x402 into developer tools and merchant services. However, the protocol is not exclusive to Coinbase; other networks are also participating, though specific breakdowns were not fully disclosed. The scale of activity signals that AI agents are already moving beyond experimental use cases, offering a tangible crypto use case that goes beyond speculation.
As USDC continues to maintain its position in the market, developments like these suggest growing institutional interest in how cryptocurrencies can facilitate AI-driven transactions. This could bolster USDC adoption as a key player in the evolving financial landscape. The announcement may also prompt regulators to examine how existing financial rules apply to transactions initiated by software rather than humans.
