In the time it took you to read this heading, your AI might have already executed multiple payments. You didn’t approve any, nor did Visa process them. According to the most fervent proponents of the crypto industry, this isn’t a flaw — it represents the fundamental future of the online economy.
Brian Armstrong, founder of Coinbase, posits that AI agents will soon surpass humans in the sheer volume of internet transactions. Binance founder Changpeng Zhao took this prediction further, suggesting that agents will conduct a million times more payments than people, exclusively using crypto. These insights were shared on the same day last week, sparking considerable discussion on crypto X.
The fundamental premise is architectural.
AI agents face a hurdle in opening traditional bank accounts due to the identity verification requirements that software cannot fulfill. In contrast, a crypto wallet merely necessitates a private key. No KYC, no regulatory checks, no waiting periods — this inherent disparity is precisely what Armstrong underscored.
However, the wallet issue only presents half of the situation. The other crucial element is economics.
Agents do not conduct transactions in the same manner as humans. When an AI agent undertakes a task—whether it’s researching a subject, orchestrating a supply chain, or compiling a report—it might make calls to dozens of specialized APIs within a single session.
Each such call could be valued at mere fractions of a cent, covering costs like GPU compute time, real-time data streams, web scraping utilities, or even engaging a sub-agent for translation services. None of these micro-transactions align with the processing capabilities Visa or Mastercard were designed for.
Imagine, for a moment, that this very article was composed by an agent, at the behest of a “chief” agent at CoinDesk whose mandate was to amplify the site’s authority.
To achieve this, the agent would have queried a real-time news API to confirm Armstrong’s tweet ($0.002), retrieved on-chain data for volume figures ($0.004), cross-referenced press releases ($0.001), and consulted a financial context model for Visa protocol specifics ($0.003). The article would then be generated at an additional expense, with credits paid to another AI tool specifically for its writing.
Based on current figures from protocols like x402, the total journalistic cost for this piece would be less than two cents, spread across six transactions.
Conversely, Stripe’s minimum processing charge for a single transaction sits around $0.30. Pushing those six payments through a traditional card network would incur costs exceeding 100 times the actual value of the payments themselves.
Following publication, a human editor overseeing the piece might then be charged by a specialized sub-agent for SEO optimization, another for conducting plagiarism checks, and yet another for formatting the content for CMS software. While each of these micropayments would be economically prohibitive on card rails, they are trivial to execute on-chain.
This principle underpins x402, Coinbase’s open payment protocol, designed to seamlessly embed stablecoin payments directly into HTTP requests. This enables an agent to encounter a paywall, pay in USDC, and proceed with its task within the same interaction, entirely without human intervention. Cloudflare, Circle, AWS, and Stripe have all endorsed this initiative. Google’s overarching standard for agent payments also incorporates x402 as a settlement layer.
Any industry characterized by high-frequency, low-value data exchanges becomes a prime candidate for this model.
In healthcare, an agent managing a patient’s insurance claim could pay for each document retrieved from a medical records API. In logistics, a procurement agent might auction freight slots across numerous carriers in real time, settling the winning bid instantly. For media, AI crawlers could pay per article indexed, sidestepping complex bulk licensing negotiations. In finance, a trading agent might pay a specialist model mere fractions of a cent for each risk signal consumed.
A significant observation, however, is that the underlying infrastructure currently outpaces the actual demand.
CoinDesk reported recently that x402 processes approximately $28,000 in daily volume. Alarmingly, Artemis identified roughly half of these observed transactions as artificial activity rather than genuine commercial exchanges. The types of merchants x402 was specifically developed to assist remain uncommon.
Meanwhile, conventional finance isn’t remaining static. Visa introduced its Trusted Agent Protocol last October, and Mastercard successfully executed Europe’s inaugural live AI-agent bank payment within Santander’s regulated framework just last week. Both initiatives leverage existing card networks, enhanced with cryptographic verification.
The most probable scenario is a bifurcation: regulated commercial transactions will likely persist on traditional card rails, while machine-to-machine payments—such as agents contracting other agents, paying per API call, or purchasing on-demand compute—will gravitate towards stablecoins, driven by economic necessity.
The ultimate question is which of these two segments will eventually dominate.