BlackRock has identified agentic artificial intelligence as a potential source of demand for stablecoins and programmable payment infrastructure, while cautioning that machine-initiated payment activity is still at an early stage.
In a September research paper titled The Machine-Native Economy, the asset manager argues that AI systems capable of planning and executing multistep tasks will need ways to make purchases, pay for data and acquire computing resources. Stablecoins could be one settlement instrument for those transactions because they combine a familiar unit of account with programmable, always-available blockchain rails.
The paper is a market thesis, not a product announcement. BlackRock does not disclose a new payment service, stablecoin, customer deployment or launch timetable. Its authors also acknowledge that agentic payment activity remains nascent and that traditional payment systems will continue to be important where AI agents interact with consumers and human-operated businesses.
Two payment models are taking shape
BlackRock distinguishes high-frequency machine-to-machine transactions from conventional commerce. It argues that blockchain-based instruments may be suited to low-value payments for API calls, data and computing capacity, while adapted card and bank-payment systems can connect agents to existing merchant checkout.
That distinction is reflected in the protocols reviewed in the paper. Coinbase’s x402 uses the HTTP 402 status code to facilitate machine-initiated payments and has stablecoins as an early use case. The Machine Payments Protocol developed by Stripe and Tempo supports settlement through stablecoins or traditional payment methods. Other initiatives focus on checkout integration, user mandates, audit trails and merchant verification rather than replacing existing payment acceptance.
For payments providers, the competitive question is therefore broader than which rail moves funds fastest. A workable agentic-payment stack must connect an agent’s instruction to a valid user mandate, verify the parties and transaction, apply spending limits and preserve records that merchants and payment providers can use when something goes wrong.
Existing rails have strengths and constraints
BlackRock says current card and ACH systems already support substantial automation. It nevertheless argues that account onboarding, credentialing and authorization can require human involvement; merchant fees may make very small payments uneconomic; and settlement timing or dispute finality can extend beyond the initial authorization.
Those constraints do not establish that stablecoins are automatically cheaper or safer. Blockchain transactions introduce their own operating decisions, including wallet security, identity and compliance controls, network selection, transaction fees, asset redemption and the treatment of errors or unauthorized transfers. The paper notes that anti-money-laundering, know-your-customer and what it calls know-your-agent checks generally occur off-chain, with verified results passed to on-chain systems.
For merchants and processors, the relevant test will be whether an agentic-payment design delivers reliable authorization, settlement and recourse across the full transaction lifecycle. Rail programmability alone does not answer who bears liability, how refunds work or how an agent’s authority is revoked.
Compute payments extend the thesis
BlackRock also describes a possible market in standardized claims on computing capacity. Under that model, capacity rights could be represented, transferred, pledged as collateral and settled through programmable infrastructure. AI agents might eventually compare providers and purchase computing resources on demand.
This part of the thesis is more prospective. The report says contract design and market structure would have to account for differences in chip performance, location and energy costs, while standards for delivery and cash settlement remain to be developed. It also says liquidity in compute markets is limited.
The immediate significance for the payments industry is not proof of a stablecoin-led machine economy. It is that a major asset manager now treats agentic commerce as a payments-infrastructure question spanning stablecoins, conventional rails, identity, mandates and settlement. Providers that want to serve software agents will need to prove the controls around those transactions, not merely automate the transfer of value.