Infrastructure bottlenecks are hindering the widespread commercial adoption of an autonomous AI agent economy, according to findings from Visa and investment thesis platform Artemis.

The Clash Between AI and Traditional Payments

A joint report published on Wednesday by the two firms highlights a significant challenge facing the incumbent global card payments infrastructure: its struggle to process the high-frequency micropayments generated by artificial intelligence (AI) agents. The report states that traditional card systems were engineered for human commerce, which is characterized by low-frequency transactions. This model is insufficient for AI agents, which demand an infrastructure capable of near-zero fees and faster settlement to make their so-called agentic micropayments commercially viable.

The need for new infrastructure has become increasingly urgent. The report notes that AI agents reached a crucial capability threshold in mid-2025, which allowed them to begin discovering unfamiliar APIs, assessing prices, and making autonomous payment decisions. While AI agents are now initiating a foundational shift in commerce, these existing infrastructure gaps are restricting their mainstream adoption.

Quantifying the Economic Potential

The scale of this emerging market was underscored earlier this week by the Australian crypto exchange Swyftx. Their analysis suggested that AI-enabled microbusinesses have the potential to drive an extra $262 billion in stablecoin volume by the year 2033. This forecast is based on an assumed adoption rate of approximately 33% for AI-native payments settled in stablecoins.

Signs of New Standards and Adoption

Despite the infrastructure challenges, some agentic payment standards are already demonstrating user adoption. One example is the x402 payment protocol, which was developed by Coinbase.

Since its launch in May 2025, the x402 protocol has processed $15 million in adjusted volume through more than 109 million adjusted transactions. The protocol experienced a dramatic acceleration in October 2025, when its monthly transaction count surged from 40,000 to 3.8 million. This momentum resulted in 38 million transactions being processed in October alone.

A Unified Path Forward

In their report, Visa and Artemis proposed that a solution could lie in a unified framework capable of handling a new type of commerce. They stated that "a single machine payment framework can support both stablecoin-based flows and card transactions, creating a path into agentic payment flows for card networks."

Such a framework would bridge the gap between emerging and traditional financial systems. The report added that Tempo’s Machine Payment Protocol (MPP) is an example of this approach, as it already covers both onchain crypto payments and fiat payments through shared payment tokens. Visa also noted that its Card Specification SDK was specifically created to extend the protocol into the domain of card-based agent commerce.

These developments align with recent industry moves. In March, both Visa’s crypto division and the Stripe-backed company Tempo launched new AI tools. Visa's tool enables AI agents to execute same-day payments. That same month, Tempo introduced its Machine Payments Protocol, designed to simplify the process for AI actors to send and receive money.