While you were reading this article, AI-driven agents executed various transactions without any approvals from you or traditional financial institutions like Visa. Prominent figures in the cryptocurrency realm, including Coinbase founder Brian Armstrong and Binance’s Changpeng Zhao, believe that soon there will be more AI agents making transactions online than there are humans, with Zhao estimating a staggering one million times more crypto transactions initiated by these agents compared to people.
Their assertions point to a significant shift in the economic landscape shaped by AI. A key argument revolves around the accessibility of crypto wallets compared to traditional bank accounts, as banks require identity verification that AI cannot provide while a crypto wallet merely requires a private key. This lack of Know Your Customer (KYC) protocols and compliance hurdles is a crucial element of attraction for AI agents seeking to transact autonomously.
However, the wallet dilemma is just one facet of a larger picture. Unlike human consumers, AI agents conduct transactions differently. When performing tasks such as research or supply chain coordination, an AI may interact with multiple specialized APIs in a single session. Each of these interactions could cost mere fractions of a cent—for instance, payments might be made for computational power, real-time data feeds, or even sub-agents tasked with translations. These microtransactions fall outside the traditional frameworks designed by payment giants like Visa or Mastercard.
To illustrate this, consider a scenario where an AI agent crafts a news report. It might query a real-time news API, cross-reference existing press releases, and gather on-chain data for volume figures, incurring costs that total under two cents across multiple microtransactions. Contrast this with traditional payment systems, where a minimum processing fee for a single transaction can reach approximately $0.30—meaning that executing the six concurrent payments through standard card networks would cost significantly more than the total transaction value.
The costs align more favorably with the x402 protocol, an open payment standard promoted by Coinbase that enables seamless stablecoin payments embedded in HTTP requests, allowing for transactions to be completed without requiring human intervention. This protocol garners support from notable firms like Cloudflare, Circle, and Stripe, and is integrated into Google’s newly developed standard for open agent payments.
Potential applications for this technology span various sectors. In healthcare, for example, an AI agent handling a patient’s insurance claim might pay small amounts for each document it accesses. Similarly, in logistics, it could auction freight slots in real time across multiple carriers, settling bids instantly. In media, AI crawlers could pay for every indexed article, moving away from conventional bulk licensing arrangements, while financial trading agents might settle costs based on risk signal consumption.
Nonetheless, despite the promise of these technologies, the demand for such infrastructures is still in development stages. Current data indicates that the x402 protocol is processing approximately $28,000 in daily volume, with reports highlighting that nearly half of these transactions may not represent genuine commercial activity.
Traditional finance is also evolving, exemplified by Visa’s recent launch of its Trusted Agent Protocol and Mastercard’s successful completion of Europe’s first AI-agent bank payment on existing card rails.
Ultimately, the future landscape may see a divergence where regulated commerce remains on traditional payment infrastructures while machine-to-machine payments increasingly migrate towards stablecoins due to the economic advantages they offer. The question that remains is which of these sectors will prove to be larger as this new ecosystem develops.
