BlackRock has put a fresh institutional spotlight on the long-running idea that artificial intelligence could translate into real demand for digital assets. In a new research paper, the worldโs largest asset manager argues that broader AI adoptionโespecially agentic, machine-to-machine activityโcould act as a structural catalyst for blockchain-based infrastructure, including stablecoins and other on-chain programmable instruments.
The report, titled โThe Machine-Native Economy,โ also points to a second potential pathway: the compute market itself. BlackRock suggests that as AI companies increasingly manage and optimize access to processing capacity, tokens could emerge as a way to represent claims on compute resourcesโpotentially enabling trading and collateral useโthereby widening the scope of what digital assets could support beyond payments.
Key takeaways
- BlackRock frames AIโparticularly autonomous agent systemsโas a โstructural catalystโ for digital asset adoption.
- The firm argues stablecoins are likely to lead transactional use for high-frequency, low-value machine-to-machine payments.
- BlackRock sees an emerging opportunity for tokenized access to AI compute, potentially used for transfer, pledging, and collateral.
- The research emphasizes the practicality gap in existing payment rails for automation tasks that require authorization, credentialing, or fast finality.
Why BlackRock thinks AI will change payment demand
BlackRockโs central payments argument is that the shift from human-led commerce to agent-driven transactions may stress existing payment systems. In the paperโs view, even where todayโs rails can support some automation, they may still require human involvement around key operational steps such as account setup, credentialing, and authorization. That friction could be amplified when transactions happen continuouslyโwithout a human operator to manage exceptions, timing, or settlement requirements.
The report also highlights economics and throughput. Merchant fees can make very small transfers uneconomic, while settlement and finality characteristics can differ across providers. For AI agents that transact around the clock and at potentially sub-cent value levels, these constraints could make traditional rails less suitable.
Against that backdrop, BlackRock argues that digital assetsโespecially stablecoinsโare better aligned with machine-to-machine needs. The paper states that stablecoins, native cryptocurrencies, and tokenized real-world assets could support high-frequency payments, but it goes further by asserting that stablecoins are the most likely to dominate actual transactional usage.
โSeveral types of digital assets may support agentic commerce, but stablecoins are likely to lead transactional use,โ the authors wrote.
Compute tokens: a second lane for digital asset demand
Beyond payments, BlackRockโs paper attempts to broaden the mental model for where blockchains could fit into an AI-driven economy. It argues that demand for computeโthe processing capacity required to train and run AI systemsโcould create a new market for digital assets.
As AI workloads grow, BlackRock suggests AI firms may try to lock in costs and manage provider risk. In such a scenario, claims on compute capacity could potentially be represented using tokens. Those tokenized claims could be transferred between parties, used as collateral, or traded, effectively turning compute access into a more modular asset class within a digital settlement layer.
The paper further proposes that AI agents could participate in these markets automatically, purchasing compute resources as needed. If that automation becomes widespread, the authors argue it could also broaden institutional participationโpositioning compute as a new opportunity for the broader digital asset ecosystem.
How this thesis echoesโand pressuresโindustry narratives
BlackRockโs framing aligns with an argument that has circulated across parts of the crypto industry: AI doesnโt replace crypto so much as it increases the need for programmable financial tools. That view has been expressed publicly by Coinbase CEO Brian Armstrong, who pushed back on calls for crypto to pivot away from its core value proposition. In July, Armstrong argued that AI is a megatrend that increases the demand for programmable money rather than traditional banking rails, implying that agentic activity could make crypto more relevant rather than less.
BlackRockโs research can be read as an institutional translation of that same themeโmoving the conversation from speculative โAI + cryptoโ narratives to specific infrastructure functions: machine-native payments and tokenized access to compute.
What existing tools suggest about agent payments
While BlackRockโs paper is forward-looking, it also arrives as crypto-native companies build tooling aimed at enabling autonomous payments. Multiple projects highlighted in earlier coverage have focused on letting AI agents automatically pay for online services or initiate recurring transfers.
For example, Cointelegraph previously noted that Coinbase introduced its x402 protocol to support agent-driven payments. Tempo has also been linked with a Machine Payments Protocol designed to support payments for machine-driven workflows. The same pattern appears in Circleโs agent wallet and USDC payment tooling, and in OKXโs work on an Agent Payments Protocol intended to support recurring payments and escrow-like arrangements released after a taskโs completion.
These products donโt prove BlackRockโs compute-token thesis, but they do show momentum around the narrower payments piece of the argumentโespecially around payments designed for autonomy, timing, and reduced human intervention.
What to watch next
If BlackRockโs thesis is directionally right, the most important developments to monitor will be real-world volumes of machine-to-machine transactions using stablecoins or other on-chain payment rails, and whether tokenized representations of compute access move from conceptual proposals into operational markets. The next step will likely determine whether โAI as infrastructure demandโ becomes a measurable adoption driverโor remains a compelling institutional hypothesis.
