analysis

BlackRock: Stablecoins Will Power AI Agent Payments

Editorial · Sep 23, 2026 · 8 min read

BlackRock has published a digital-assets research paper arguing that stablecoins, not bank rails, will be the money AI agents use to pay each other. The claim is straightforward: agents buying data, software and compute need continuous, programmable, low-friction settlement, and cards and wires were never built for that. When the world’s largest asset manager puts machine-to-machine stablecoin payments in writing, it stops being a crypto-native thesis and becomes an allocation question.

What the Paper Actually Argues

The core of BlackRock’s argument is a mismatch between payment infrastructure and machine behavior. Credit cards settle in batches, charge high fixed fees per transaction, and assume a human cardholder behind an authorization. Wire transfers are slower and more expensive still. An AI agent that needs to pay a few cents for an API call, a data feed, or a minute of GPU time cannot use either rail economically.

Stablecoins solve this because they are programmable, divisible to arbitrary precision, and settle in seconds on public chains. The paper frames agents as a genuinely new source of stablecoin demand: not people moving dollars on-chain, but machines paying machines continuously for inputs. That is a volume profile that behaves more like infrastructure metering than like payments as traditionally counted.

Why Ethereum and Circle’s Arc Get Named

Commentary around the paper notes that BlackRock points to Ethereum and Circle’s Arc as likely beneficiaries. That is not a neutral observation. BlackRock issues BUIDL on Ethereum and has a deep commercial relationship with Circle, whose Arc blockchain is purpose-built for stablecoin settlement. The research thesis and the firm’s own infrastructure positions point in the same direction.

Still, the underlying logic holds independently of the conflicts: agent payment traffic values cheap, fast, reliable finality and deep stablecoin liquidity, and those properties concentrate on a small number of chains. Ethereum’s L2 ecosystem and Circle’s purpose-built settlement layer are the obvious candidates today, though the x402 implementations spreading across other chains compete for the same traffic.

What This Means for Stablecoin Demand

The interesting number is not today’s stablecoin float but the transaction count agents could generate. Human retail payments are episodic — a purchase here, a remittance there. Agent payments are continuous: an agent running a workload might execute hundreds of micro-settlements per hour across data vendors, compute markets and software services. Per-transaction value is tiny; aggregate volume could dwarf human-driven flows.

For issuers, this changes the reserve economics conversation. Float held for agent working capital behaves differently from float held for speculation or savings — it turns over faster and is less sensitive to yield elsewhere. BlackRock’s framing also implicitly supports the facilitator and gas-abstraction models that Circle and others have been shipping, since agents cannot practically manage gas tokens and screening themselves.

The Skeptical Read

There are open questions the paper’s headlines skip over. First, agent payment volume today is negligible relative to stablecoin float; the thesis is a forecast, not a measurement. Second, the incumbent banks are not standing still — six major banks published joint agentic-commerce principles this week, and card networks are building agent rails of their own. Whether agents settle on stablecoins or on card-adjacent infrastructure is a live contest, not a settled conclusion. Third, identity, liability and spending-limit enforcement for autonomous agents remain unsolved; a BlackRock paper does not fix them. Still, the direction of travel is clear: the largest asset manager now publicly expects machines to hold the money it helps tokenize.

Sources

E
Editorial
Related reading

Related reading