The rapid growth of automated payments between AI agents is creating demand for a new class of payment solutions designed to process millions of transactions worth less than $1.
Payment network Visa and on-chain data analytics firm Artemis published a joint report titled Agentic Payments from the Ground Up. The report examines the development of infrastructure that allows AI agents to pay autonomously for digital services, computing resources, and data access.
As part of the research, the authors analyzed two open protocols designed for automated payments between AI agents:
- x402, originally developed by Coinbase and Cloudflare before its governance was transferred to the Linux Foundation, processed 109.6 million transactions worth about $15 million since its launch in May 2025.
- Machine Payments Protocol (MPP), developed by Stripe and Tempo with contributions from Visa, processed about 115,000 transactions worth approximately $25,000 during the first weeks following its launch in March 2026.
The analysts found that the average payment size on both networks amounted to only fractions of a cent. As a result, fees charged by traditional card networks often exceeded the value of the transaction itself, making existing payment infrastructure inefficient for these use cases.
The researchers identified two primary segments of agentic payments:
- Large-value purchases, where an AI agent acts on behalf of a user by managing subscriptions, booking travel, or purchasing goods. These transactions closely resemble traditional e-commerce and can rely on existing card infrastructure.
- Micropayments between software systems, where one system automatically pays another for access to data, computing resources, or individual requests.
According to Visa and Artemis, the second segment requires a new payment infrastructure because existing systems aren’t designed for the rapidly expanding market of autonomous AI agents.
A separate section of the report focuses on trust and regulation. The authors noted that today’s dispute resolution mechanisms and liability frameworks were developed for payments initiated by humans. They don’t address scenarios in which autonomous AI agents execute thousands of transactions per hour.
The analysts identified several unresolved issues, including:
- assigning liability for erroneous purchases;
- protecting AI agents from external attacks and fraud;
- establishing unified digital identity standards for autonomous software.
At the same time, the analysts said the growth of the agentic payments market won’t trigger competition between payment cards and stablecoins. According to the report, payment cards will remain the preferred option for higher value purchases made by AI agents on behalf of users, while stablecoins and blockchain infrastructure will be better suited for machine-to-machine micropayments that require minimal fees and near instant settlement.
Earlier, the analysts said automated AI agents could fundamentally reshape the micropayments market. At the same time, officials at European central banks and financial regulators believe advances in AI technology are creating new risks for financial markets.
