Agents Are Set to Pay for Everything
The idea that future money will be handled by computer programs instead of people has gained traction among crypto enthusiasts. Two influential founders, Brian Armstrong of Coinbase and Changpeng Zhao of Binance, recently shared that they expect autonomous agents to outnumber humans in online transactions and that these agents will rely on cryptocurrencies. Their comments sparked lively discussion on crypto‑focused social media.
The main point is that software can open a digital wallet with just a private key, whereas banks demand identity checks that machines cannot perform. This gives agents a clear advantage for making small payments, especially when the amounts are so tiny that traditional card processors would charge a fee that exceeds the value of the transaction. For instance, an AI could call dozens of specialized services—like data feeds or translation tools—in a single task. Each request might cost fractions of a cent, and the total could stay below two cents using stablecoin protocols.
An example of this micro‑payment model is a hypothetical article written by an AI for a news site. The bot might use separate APIs to fetch real‑time data, check on‑chain activity, and compare press releases. Each step would be paid for in stablecoin, with the final cost remaining well under $0.02. In contrast, a traditional credit‑card route would charge around $0.30 per transaction, making the same work far more expensive.
This approach is embodied in x402, a protocol that lets agents embed stablecoin payments directly into their HTTP requests. Big names such as Cloudflare, Circle, AWS, and Stripe support it, and Google has included x402 in its own open‑agent payment standard. The promise is that a machine can pay for services, retrieve data, or even hire another sub‑agent without any human intervention.
Industries that involve rapid, low‑value exchanges could benefit most. In healthcare, a bot might pay for each medical record it pulls; in logistics, an agent could bid on freight slots and settle instantly; media crawlers might pay per article indexed rather than negotiate large licensing deals; and trading bots could buy risk signals for a few cents each.
However, the system is still early. Recent reports show that x402 handles only about $28,000 of daily volume, and many of those transactions are flagged as artificial rather than genuine commerce. Meanwhile, traditional payment networks are catching up: Visa has introduced a Trusted Agent Protocol and Mastercard recently completed an AI‑driven bank payment in Europe, both using existing card infrastructure.
The likely future is a mix. Human‑led commerce will probably stay on familiar card rails, while machine‑to‑machine payments—like agents hiring other agents or buying compute on demand—will shift to stablecoins because the economics favor it. The question remains: which side will grow larger?
