According to a recent report from Keyrock, stablecoins operating on blockchain networks are becoming the primary payment method for AI agents, as traditional card systems struggle with micropayment transactions. Key insights include:

— AI agents have processed over $73 million through 176 million blockchain transactions in the past year, as per Keyrock, although this represents only a small portion of the global payments landscape.

— Major companies like Coinbase, Stripe, Google, and Visa are developing competing infrastructures for machine-to-machine payments, as software agents are increasingly autonomously acquiring data, computing resources, and digital services.

— Presently, nearly all transactions by agents are conducted in USDC, underscoring Circle’s rising significance in the crypto payment sphere and the risks associated with relying on a single stablecoin provider.

The market for AI agents spending autonomously online is still in its infancy, yet some of the largest technology, payment, and cryptocurrency companies are already in a race to establish the necessary infrastructure, as highlighted in Keyrock’s report.

The crypto trading and investment firm projects that AI agents executed over $73 million in approximately 176 million blockchain transactions from May 2025 to April 2026. While these figures are minor compared to traditional finance (TradFi)—with Visa alone processing $14.5 trillion each year—the focus lies less on the total U.S. dollar amount and more on the rapid establishment of the infrastructure, according to the report. Global entities such as Coinbase (COIN), Stripe, Google (GOOG), and Visa (V) have all introduced competing systems for machine-to-machine transactions.

The broader concept of agentic payments suggests that software increasingly consumes digital services on its own rather than relying on human-managed subscriptions and accounts. For instance, an AI trading agent could continuously acquire market data, cloud computing, or AI-generated analysis in small increments throughout the day without needing human approval for each transaction.

This potential is fueling ambitious predictions regarding the growth of the agentic payment sector. Gartner anticipates that AI agents could facilitate $15 trillion in purchases by 2028, while McKinsey estimates that retail agentic commerce could reach between $3 trillion and $5 trillion by 2030, as per the Keyrock report.

These forecasts suggest growth rates that may outpace those experienced by stablecoins during their initial breakout years, according to the report, which also indicates that the current pace of infrastructure deployment signals that the market is moving beyond the experimental stage.

Coinbase’s x402 protocol has emerged as a leading crypto-native solution, enabling AI agents to directly pay with USDC for services such as blockchain analytics or cloud infrastructure without the need for account creation or subscriptions. Stripe has launched a competing framework called Machine Payments Protocol (MPP) through its Tempo blockchain, while Google has introduced AP2, a system tailored for delegated spending authorization for AI agents. Visa has expanded its card network with tokenized credentials aimed at AI-driven commerce.

Crypto networks and stablecoins are becoming the preferred settlement layer, and the economics illustrate why.

According to the report, around 76% of agent transactions are below the 30-cent fixed fee commonly associated with card payments. Most payments fall between one and ten cents, making traditional payment systems impractical for automated software agents purchasing data, AI inference, or API access. In contrast, stablecoin transactions on certain blockchains like Base and Tempo incur costs of mere fractions of a cent.

Currently, 98.6% of machine payments are conducted using USDC, the stablecoin issued by Circle (CRCL). This reinforces Circle’s role in crypto payments, but it also raises concerns about concentration risk, creating a reliance on a single issuer.

Regulatory measures may pose challenges to growth. The MiCA regulation in Europe, the U.S. GENIUS Act, and the EU AI Act are all expected to come into effect around mid-2026, but none directly address autonomous machine-to-machine transactions or issues surrounding liability and agent identity, as noted in the report.