BlackRock Paper Backs Stablecoins as Rails for AI Agent Payments

BlackRock has concluded that stablecoins are better suited than banks or card networks to the sub-cent, round-the-clock payments that autonomous AI agents would need to make, according to a paper reported by Decrypt. The asset manager does not claim a finished system. CoinDesk reported that the firm says agent payments remain at an early stage, and that liquid markets for standardized compute contracts have yet to develop.

The caveat matters more than the headline. BlackRock highlights Coinbase's x402 protocol as one emerging way for agents to pay for online resources, including API calls, yet the on-chain evidence for genuine agent activity remains thin.

TRM Labs, a blockchain analytics firm, examined $52.7 million in x402 settlements. The firm estimated that AI agents likely accounted for between 0.6% and 7.5% of that volume, InvestingLive reported. Even at the top of that range, machines drove less than $4 million — roughly the annual budget of a mid-sized engineering team, not a new market structure.

The takeaway: the infrastructure argument is running well ahead of measured demand.

What do agents actually need?

Executives from Coinbase and Chainlink, speaking at a BlackRock forum, outlined their requirements. Coinbase's leadership, represented by Lovett, said agents need new low-cost rails built for very high volumes. He said Coinbase is investing heavily in stablecoin payments, fast blockchains and agent protocols. No dollar figures for that investment were disclosed.

Chainlink co-founder Sergey Nazarov expects tightly controlled use of agents. In his view, an agent would carry an identity and operate within set spending limits — a framework analogous to a Know Your Agent check. TokenPost reported that the discussion centered on trusted data, guardrails and verifiable execution for agentic transactions.

Chainlink operates a decentralized oracle network that connects smart contracts with external data sources, APIs and payment systems, which places it directly in the path of any machine-to-machine payment flow that requires off-chain data.

The takeaway: identity and spending limits, not raw throughput, are the stated prerequisites.

Why traditional rails fall short

The BlackRock paper's core claim is architectural. Card networks and bank transfers were designed for human-scale transactions: batch settlement, business-hours cutoffs and fees that exceed the value of a single API call. Agents buying compute or data in micro-increments, at all hours, need sub-cent pricing that legacy rails cannot clear economically.

Stablecoins settle on blockchains around the clock and can carry programmable limits. That combination, not yield or speculation, is what the paper identifies as the fit. The claim is plausible on its own terms. It is also untested at scale, since the TRM data suggests agent-driven volume is a rounding error today.

The takeaway: the technology fits the use case; the use case has yet to arrive in size.

What comes next

The panel reflects a wider push by traditional finance and crypto firms to define the infrastructure for software-driven payments. For now, the debate centers on cost, speed and control, with identity and spending limits seen as prerequisites for wider use.

Watch for two signals over the coming quarters: growth in the share of x402 settlements attributable to agents, and the first standardized compute contracts that would give agents something liquid to buy. Until both appear, the demand wave BlackRock describes remains a forecast, not a fact.

By Elena Vasquez, Mempool Brief