Artificial intelligence (AI) agents autonomously spending money online is still a small market, but some of the world’s biggest technology, payments and crypto companies are already building the infrastructure for it, Keyrock said in a new report.
The crypto trading and investment firm estimated that AI agents settled over $73 million across about 176 million transactions on blockchain rails between May 2025 and April 2026.
The amounts remain insignificant compared to traditional financing (TradFi). Visa, for example, alone processes $14.5 trillion annually. But the significance lies less in the overall US dollar value and more in how fast the infrastructure stack is forming, the report argued. Global firms such as Coinbase (COIN), Stripe, Google (GOOG) and Visa (V) all rolled out competing systems for machine-to-machine payments.
The broader idea behind agent payments is that software increasingly consumes digital services autonomously rather than through human-managed subscriptions and accounts. For example, an AI trading agent could continuously purchase market data, cloud computing or AI-generated analytics in small intervals throughout the day without a human manually authorizing each payment.
That potential is driving ambitious forecasts for how big the agent payments sector could grow. Gartner projects that AI agents could mediate $15 trillion in purchases by 2028, while McKinsey estimates retail sales could reach $3 trillion-$5 trillion by 2030, according to the Keyrock report.
Those projections imply growth rates even faster than stablecoins experienced during their breakout years, the report said, but said the pace of infrastructure deployment already signals the market is moving beyond its experimentation phase.
Coinbase’s x402 protocol has emerged as one of the leading crypto-native systems. The protocol allows AI agents to pay directly with USDC for services such as blockchain analytics or cloud infrastructure without creating accounts or subscriptions.
Stripe, with its Tempo blockchain, launched a competing framework called Machine Payments Protocol (MPP), while Google introduced AP2, a system focused on delegated spending authorization for AI agents. Visa has expanded its card network with tokenized credentials designed for AI-powered commerce.
Cryptostrips and stablecoins are emerging as the settlement layer of choice, and the economics help explain why.
About 76% of agent transactions, according to the report, fall below the flat 30-cent fee common to card payments. Most payments ranged between one and 10 cents, making traditional rails impractical for automated software agents buying data, AI inference or API access. Meanwhile, stablecoin settlement on some blockchains like Base and Tempo costs fractions of a cent.
Currently, 98.6% of machine payments are settled in USDC, the stablecoin issued by Circle (CRCL). It strengthens Circle’s position in crypto payments, but also introduces the risk of concentration, creating dependence on a single issuer.
Regulation can be a source of limitation to growth. MiCA in Europe, the US GENIUS Act and the EU AI Act are all expected to enter into force around mid-2026, but none of them directly address autonomous machine-to-machine transactions or issues around liability and agent identity, the report noted.
