Keyrock Tallies $73 Million in AI Payments
Keyrock says AI agents have settled more than $73 million across 176 million transactions, offering the clearest evidence yet that machine-to-machine payments are operating at meaningful scale and that traditional card rails may not be built to handle them.
The figures come from a new Keyrock report titled “Who Pays the Agent?”, which maps out how autonomous software agents are generating real payment volume. The report found that 76% of that activity fell below the $0.30 card-fee floor, with the median transaction sitting between $0.01 and $0.10.
That fee-floor detail matters. At those transaction sizes, Visa and Mastercard minimum fees would consume the entire payment, making traditional card networks economically unworkable for the bulk of machine-initiated commerce.
USDC Dominates Machine Payments
Keyrock’s data shows that 98.6% of the 176 million payments settled in USDC, making Circle’s stablecoin the de facto settlement layer for AI agent transactions. USDC currently holds a market cap above $76.4 billion.
The stablecoin’s dominance in this context is not surprising. Agents need programmable, instant-settlement currency that works across chains without conversion friction. USDC offers that with broad exchange and protocol support, a dynamic that echoes the broader trend of crypto rails becoming default infrastructure for new financial flows.
Four Competing Architectures, Big-Name Backers
The report identifies four distinct machine-payment architectures now in development or production. Coinbase, Stripe, Google, Visa, and American Express are all building competing infrastructure around the same trend.
One of those architectures, x402, already shows live traction. The protocol’s own site displays 75.41 million transactions and $24.24 million in volume over the last 30 days alone, suggesting the pace of adoption may be accelerating beyond what the 12-month tally captures.
Google’s approach, called AP2, takes a different angle. Its protocol uses typed mandates, configurable guardrails, and a payment receipt audit trail for agent spending. The model is designed to give businesses control over what an agent can authorize and how much it can spend, an approach closer to corporate expense management than open payment rails.
According to Keyrock’s report, incumbents have deployed more than $8 billion in acquisitions to build machine-payment stack coverage, though that tally has not been independently audited.
Why Traditional Rails Cannot Serve This Market
The core tension in AI payments is scale versus unit economics. When three-quarters of transactions cost less than a dime, fixed-fee payment systems are structurally excluded. This is not a marginal inefficiency; it is a fundamental incompatibility.
Crypto-native rails solve this by settling on-chain at near-zero marginal cost. The combination of stablecoin settlement and protocol-level micropayment support creates infrastructure that card networks were never designed to replicate. For traders and builders watching how crypto infrastructure projects gain traction, this is a category worth tracking.
Keyrock also reported that AI agents account for 37% of all Safe transactions on Gnosis Chain and over 75% on peak days, according to unconfirmed figures from the same report. If accurate, that would make agent activity a dominant force on at least one major chain.
Regulation Has Not Caught Up
Keyrock notes that MiCA, the U.S. GENIUS Act, and the EU AI Act are converging in mid-2026, but none of these frameworks directly address autonomous machine-to-machine transactions. Liability, authorization, and agent identity remain unresolved.
That gap creates both opportunity and risk. Protocols like x402 and AP2 are building trust layers now, but without regulatory clarity, the rules governing who is responsible when an agent overspends or mispays remain undefined. The emergence of new contract structures in crypto markets shows how quickly infrastructure can outpace the rules designed to govern it.
What to Watch Next
The $73 million figure is a 12-month snapshot, but x402’s 30-day run rate suggests the trajectory may be steepening. Whether that translates into sustained growth depends on several factors: regulatory clarity, stablecoin adoption in non-crypto contexts, and whether agent-payment protocols can move beyond developer tooling into production enterprise use.
The Fear and Greed Index currently sits at 25, reflecting extreme fear across crypto markets. That risk-off backdrop makes the infrastructure buildout story easier to overlook, but the transaction volumes suggest machine payments are growing regardless of market sentiment.
Disclaimer: This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency and digital asset markets carry significant risk. Always do your own research before making any investment decisions.
