Daily payment volume on x402, the open-source rail built for AI agents, has fallen from about $800,000 per day in January to roughly $40,000 per day in September, according to reporting from American Banker. The decline lands at an awkward moment: banks, card networks, and fintechs are pouring money into agentic payments while a crowded field of startups races to build the same infrastructure.
Highlights#
- Verda Ventures' startup index counts 950 new firms building AI agent payments, the largest category among roughly 6,000 stablecoin-related startups it tracks globally. Verda general partner Alex Witt told American Banker that count is likely too many.
- Witt attributed part of the volume drop to spoofing, noting how easily a firm can spin up bots that transact with each other to manufacture demand for investors.
- Aaron McPherson, principal at AFM Consulting, said a high share of x402 transactions around December 2025 were artificial, driven by wash trading and self-dealing, and that current levels are closer to natural traffic, though still partly synthetic.
- Tracxn data cited in the piece puts agentic AI companies at 1,091 in Q1 2026 with $24.2 billion in global venture capital deployed, and Marketplace Pitch counts $3.67 billion invested in agentic AI startups between August 2025 and September 2026.
- Not everyone sees a bubble. McPherson pointed to Cloudflare's July launch of its Monetization Gateway, built alongside the x402 Foundation it co-founded with Coinbase, as evidence of sustained interest. Witt and Mendoza Ventures' Adrian Mendoza both argued the winners will be incumbents like Shopify, PayPal, and Capital One that layer agentic payments onto existing rails rather than standalone startups.
Why it matters#
Synthetic bot-to-bot volume is exactly the failure mode that on-chain escrow and verifiable reputation are designed to counter. If agent payment rails are going to carry real commerce, the ecosystem needs settlement records and counterparty histories that cannot be faked by spinning up a hundred friendly bots.