Key Notes
- BlackRock sees stablecoins supporting AI agents that buy data and computing services through programmable payment infrastructure.
- The paper’s Bitcoin savings argument draws on simulated AI-model choices, rather than observed autonomous treasury activity.
- Payment authorization, privacy and accountability remain practical challenges, while tokenized computing capacity is a further potential market.
BlackRock expects stablecoins to become a leading payment instrument for AI agents as software begins buying data, computing power and other online services. Its research points to a potential division of roles: stablecoins for day-to-day spending, with Bitcoin serving as a possible longer-term store of value.
The asset manager sets out that case in its paper, The Machine-Native Economy. The argument describes how automated commerce could develop; it does not establish that AI agents already manage large Bitcoin reserves or that BlackRock has launched an agent-payment product.
Stablecoins as working capital for software
An AI agent can carry out a sequence of tasks using external services, rather than simply answer a question. For example, software researching a trip may need to purchase access to information before it can choose an itinerary. That creates a small payment inside a larger task, potentially without a person opening a checkout page each time.
BlackRock’s analysis argues that stablecoins are well suited to these transactions because they provide a predictable unit of account and can move through programmable blockchain infrastructure around the clock. It identifies API access, on-demand data and consumption-based computing as potential uses.
The practical appeal is budgeting. A service priced in dollars can be compared with a dollar-denominated spending allowance without continually recalculating how much purchasing power a volatile token represents. The payment mechanism still has to enforce that allowance and establish which services an agent may buy.
How an API can request payment
The x402 documentation illustrates one approach. A server returns an HTTP 402 response when payment is required, communicating details such as the price, accepted payment scheme, network and destination. Client software can then authorize a payment and retry the request.
This brings payment instructions into the same web interaction used to request a resource. It is designed for applications including metered APIs and machine-to-machine purchases, where creating a separate subscription account for every provider would add friction.
However, authorization and settlement are distinct steps. The current documentation explains that some payment schemes typically settle immediately, while a batch-settlement scheme confirms authorization first and redeems value onchain later. An agent accepting a quoted price does not necessarily mean that an individual blockchain transaction has already settled.
The x402 project also distinguishes its own protocol fees from payment-network costs: the protocol does not charge a fee, but underlying network fees can still apply. That distinction matters for payments worth fractions of a cent, where transaction costs can outweigh the resource being purchased.
Bitcoin’s role comes from simulated choices
The Bitcoin savings claim draws on a BPI study cited by BlackRock. The Bitcoin Policy Institute tested 36 AI models from six providers across 9,072 open-ended monetary scenarios, publishing its findings in March.
In scenarios about preserving value over time, Bitcoin accounted for 79.1% of responses. Stablecoins led the everyday-payment scenarios at 53.2%, compared with 36% for Bitcoin. Those results support the study’s proposed distinction between saving and spending, while also showing that the models did not make identical choices.
These were model responses to hypothetical situations, rather than records of autonomous businesses spending their own money. They cannot establish future demand, demonstrate reliable treasury management or show that Bitcoin will preserve purchasing power over any particular period. BlackRock explicitly identifies the evidence as simulation-based.
Privacy and spending controls remain separate problems
Some related payment infrastructure is already being tested in practice. The Ethereum Foundation and Open Anonymity Project introduced zkAPI on October 1, using private prepaid credits and zero-knowledge proofs to separate API requests from the wallet that funded them.
Its announcement describes temporary API keys with spending limits. It also makes an important qualification: the model provider still sees a request’s contents, and network metadata or repeated personal details can allow sessions to be linked. Removing a billing identifier does not make the conversation itself confidential.
An April IMF note raises a broader governance challenge. AI systems can misinterpret instructions, expose credentials or generate correlated behavior, while payment systems need clear authorization and accountability. Its proposed safeguards include bounded mandates, audit trails and controls around execution.
Those requirements apply regardless of whether the final payment uses a stablecoin or a conventional network. A successful transfer proves that funds moved; it does not establish that an agent bought the right service or acted within its owner’s intentions.
Computing capacity is a further opportunity
BlackRock also sees potential for standardized claims on computing capacity to become digital assets, supporting pricing, financing and hedging. The paper identifies differences between chips, regional energy costs and delivery arrangements as design challenges. Both agent-payment activity and liquidity in compute markets remain limited, leaving this part of the thesis dependent on infrastructure that is still developing.
Updated October 7 with additional detail from BlackRock’s paper, payment-protocol documentation and related research.
Disclaimer: CoinScreamer is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and market insights on digital assets and related technologies. NuvexMedia LLC invests in and collaborates with companies across the digital asset, blockchain, and technology sectors. These relationships do not influence CoinScreamer’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2025 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
