BlackRock Sees Stablecoins Powering AI Agent Payments

BlackRock says stablecoins are likely to lead digital-asset payments by autonomous AI agents, arguing that frequent, low-value purchases of data and computing power could become a new source of demand for digital assets.

In its The Machine-Native Economy research paper, the asset manager identifies an opening for programmable payments as AI systems move beyond generating content to executing tasks with limited human intervention. Stablecoins' price stability, it argues, makes them useful for pricing and settling those purchases.

BlackRock shared the paper on Sept. 22, saying broad AI adoption “may drive new demand, utility and applications across the digital asset economy.”

The proposed advantage is specific: payments for individual data requests or units of computing power that may be too small to justify conventional merchant fees. BlackRock acknowledges that bank transfers through ACH and card networks already support substantial automation, but says onboarding requirements and settlement economics can make them less suited to very low-value, always-on transactions.

From Subscriptions to Per-Request Payments

The paper highlights Coinbase's x402, an open payment protocol that uses the web's HTTP 402 “Payment Required” response. Its technical paper describes a flow in which an agent requests a resource, receives pricing and payment details, then retries with a signed payment authorization. The service verifies and broadcasts the payment before returning the requested response.

That design lets a provider charge per request instead of requiring a subscription. A concrete example came in June, when 0x announced that agents could call its Swap API and “pay $0.01 per request in USDC from their own wallet,” with no API key required. The Defiant covered the integration, built with Alchemy's AgentPay.

BlackRock also sees a possible benefit for the blockchains settling those payments: greater activity could increase demand for blockspace and validator services. It cautions that the extent of value captured by native cryptoassets depends on each network's fees, staking rules and gas-sponsorship design.

The paper does not envision every agent purchase moving onto crypto rails. It says modified traditional systems will remain important when agents transact with human-operated businesses and consumers. The Agentic Commerce Protocol's documentation, for example, describes agent-assisted checkout that lets sellers retain their existing payment processing.

BlackRock's thesis remains forward-looking. The paper describes agentic payment activity as nascent and liquidity in compute markets as limited, even as it outlines how agents could eventually compare, provision and pay for computing resources with little human intervention.