AI agents are beginning to make autonomous purchases, but measuring their spending reveals a massive gap between reported figures and verified transactions. Bloomberg reported $24 million in AI agent payments over a 30-day period, while onchain data showed roughly $3 million for the same timeframe. When researchers filtered out wash trades (artificial transactions between the same wallets), the number dropped to $1.6 million.

Why Are the Numbers So Different?

The discrepancy between $24 million and $1.6 million tells a revealing story about how immature the measurement infrastructure still is. Most AI agent activity centers on developer tools where pay-per-query models are replacing traditional subscriptions. Firecrawl charges $0.01 per web scraping query, Browserbase sells browser sessions, and Freepik offers image generation, all accepting cryptocurrency payments through the x402 protocol.

The x402 protocol, built by Coinbase, allows developers and AI agents to pay for services without creating accounts or committing to subscriptions. A developer can try a tool once with just a wallet. This frictionless payment model is expanding beyond developer tools into enterprise use cases. Merit Systems built StableEnrich, which aggregates data from Apollo, Google Maps, and Whitepages behind a single x402 endpoint, with each request costing pennies. Sales teams use AI agents to call these endpoints automatically and pay for leads.

What Are Major Tech Companies Betting On?

The infrastructure being built around AI agent payments suggests that major technology companies see far larger numbers ahead. Coinbase built the x402 protocol, and Stripe, Cloudflare, and Vercel have all integrated it into their platforms. Google embedded x402 into its agent payments protocol. None of these companies are betting their strategies on $1.6 million per month in current transaction volume.

• Current Activity: Most AI agent payments are concentrated in developer tools and data aggregation services where pay-per-query models eliminate friction.
• Infrastructure Expansion: Major platforms including Stripe, Cloudflare, Vercel, and Google have integrated or built support for AI agent payments, signaling confidence in future growth.
• Enterprise Adoption: Companies like Merit Systems are building products that let sales teams deploy AI agents to automatically purchase data and leads through these payment rails.

The infrastructure investments suggest these companies are positioning for a future where agents become the default buyer. The current $1.6 million monthly figure represents early-stage adoption, primarily among developers and technical teams. As AI agents become more autonomous and widespread across enterprise workflows, the payment volume could expand dramatically.

"The infrastructure being built around it is" what matters most, noted Noah Levine, an investment partner at a16z crypto, emphasizing that the current transaction volume understates the strategic importance of these payment systems.

Noah Levine, Investment Partner at a16z Crypto

How to Understand AI Agent Payment Infrastructure

• Measurement Challenge: Reported AI agent payments vary wildly depending on the data source and methodology, with wash trade filtering reducing estimates by over 93 percent in some cases.
• Pay-Per-Query Model: Instead of subscriptions, AI agents now pay tiny amounts per transaction, eliminating account creation and commitment friction for both agents and service providers.
• Enterprise Integration: Companies are building products that let AI agents automatically purchase data, leads, and services from multiple providers through unified payment endpoints.
• Strategic Positioning: Major technology platforms are integrating AI agent payment infrastructure despite current volumes being modest, indicating confidence in future adoption and scale.

The gap between reported and verified AI agent payments reveals an industry still in its infancy. The measurement infrastructure itself is unreliable, with different data sources producing estimates that differ by over 1,400 percent. Yet the fact that Coinbase, Google, Stripe, Cloudflare, and Vercel are all building or integrating support for these payments suggests they see a future where AI agents routinely make autonomous purchasing decisions across enterprise workflows.

For now, the real story is not the current transaction volume. It is the infrastructure being built in anticipation of what happens when agents become the default buyer rather than the exception.