When you read this headline, your AI has just made a few payments. You didn’t approve any of them. No visa has been applied for. If the cryptocurrency industry’s biggest bulls are right, then this isn’t a mistake — it’s the entire future of the internet economy.
Coinbase founder Brian Armstrong believes that soon there will be more artificial intelligence agents conducting transactions on the internet than humans. Binance founder Changpeng Zhao went a step further, predicting that agents will be paid a million times more than humans, all in cryptocurrency. The posts were published on the same day last week and ignited CryptoX.
The core argument is structural.
AI agents cannot open bank accounts because banks require authentication that software cannot provide, while crypto wallets only require private keys. No KYC, no compliance checks, no waiting – this asymmetry is exactly what Armstrong pointed out.
But wallet issues are only half the problem. The other half is economics.
Agents don’t shop like humans. When an AI agent performs a task (such as researching a topic, coordinating a supply chain, building a report), it may call dozens of specialized APIs in a single session.
Each call can be worth less than a penny, requiring payment for GPU compute time, a live data source, a web scraping service, or hiring a subagent to handle the translation. These transactions are not designed to be processed by Visa or MasterCard.
Consider for a moment that this story was written by an agent, at the request of CoinDesk’s “lead” agent, tasked with boosting the site’s authority.
To generate it, the agent will query the real-time news API to verify Armstrong’s tweets ($0.002), pull on-chain data to search for transaction volume data ($0.004), cross-reference press releases ($0.001), and ping a financial context model for Visa protocol details ($0.003). It ends up generating articles at an additional cost, paying points to another AI tool to actually Write This piece.
Based on current data provided by protocols such as x402, six transactions have a total reported cost of less than 2 cents.
In comparison, Stripe’s minimum processing fee for a single transaction is about $0.30. The cost of running these six payments through the card network would be more than 100 times the value of the payments themselves.
The human editor who reviewed and published the article may then be billed by a subagency that does SEO optimization, another subagency that runs plagiarism checks, and another that does formatting for the CMS software. Every micropayment is economically ridiculous on the rails, but trivial on the chain.
This is the argument behind x402, Coinbase’s open payments protocol that embeds stablecoin payments directly into HTTP requests so that agents can go behind a paywall, pay in USDC, and continue their tasks within the same interaction, without human intervention. Cloudflare, Circle, AWS and Stripe all support it. Google’s open proxy payment standard includes x402 as a billing layer.
Every industry with high frequency and low value of data exchange is a candidate.
In healthcare, agents who manage patient insurance claims make payments based on documents retrieved from the Medical Records API. In the logistics field, procurement agents auction freight slots from dozens of carriers in real time, settling winning bidders immediately. In media, AI crawlers are paid per indexed article rather than negotiating volume licensing deals. In finance, trading agents pay expert models a fraction of a cent for each risk signal consumed.
However, it is important to note that infrastructure lags behind demand.
CoinDesk reported this week that x402 is currently processing about $28,000 in daily transaction volume, with Artemis labeling about half of the transactions observed as human activity rather than true commercial activity. Merchants x402 built specifically for service are still rare.
Meanwhile, traditional finance isn’t standing still. Visa launched its Trusted Agent Protocol last October, and Mastercard last week completed Europe’s first real-time AI agent banking payments within Santander’s regulated infrastructure – both on existing card rails with cryptographic verification on top.
The most likely outcome is fragmentation, with regulated commerce remaining stuck on the rails, while machine-to-machine payments — such as agents hiring agents, pay-per-API calls, and purchasing compute on demand — migrate to stablecoins because the economy demands it.
The open question is which barrel ends up being bigger.