**BlackRock** said that AI agents will create new demand for stablecoins and blockchain, and that tokenized computing power will emerge as a new asset class alongside a growing cloud market projected to expand to a $1.1 trillion scale by January 2030.
Key point
BlackRock believes AI agents need “machine-native” payment infrastructure, and that stablecoins will lead in this area.
It expects that rights to AI compute (computing) capacity could be tokenized as tradable digital assets.
However, it warns that agent-based payments and the computing market are still in the early stages.
BlackRock’s stablecoin thesis
BlackRock, the world’s largest asset manager, presented this analysis in a white paper titled *The Machine-Native Economy* (machine-native economy), co-authored by Will Su, Robert Mitchnick, Jay Jacobs, and **William Helm**. They argued that AI agents that plan and carry out tasks while minimizing human involvement need not existing payment infrastructure designed for people, but a new payment method centered on machines—and that stablecoins will lead the way.
The report said that card networks and bank transfer systems have already absorbed a considerable degree of automation, but that opening accounts still requires human involvement, and that very small payments are less economical due to merchant card fees. Another weakness cited is that it takes longer to settle card payments and handle disputes.
According to BlackRock, as of September, the supply of issued stablecoins exceeded $300 billion. Under the 2025 adjusted (stablecoin-adjusted) basis, the volume of stablecoin payments surpassed $1.1 trillion, reaching a level comparable to the annual transaction volume of **Visa** and Mastercard. However, this still leaves a wide gap compared with $9.3 trillion in fund transfers via the U.S. automated clearing house (ACH) network. Even so, BlackRock assessed that the growth momentum is completely different: from 2020 to 2025, the stablecoin transaction volume grew at an average annual rate of about 80%, while ACH grew only 8.5% per year over the same period.
Reference article: Dogecoin: Surpasses all major tokens despite pressure from short-selling forces
Tokenizing the AI computing market
The authors say they see a second opportunity in the computing power used to train and run AI systems. According to analyst estimates cited in the report, the combined revenue of Amazon Web Services, Microsoft’s Intelligent Cloud segment, and **Google Cloud** is expected to reach about $1.1 trillion by 2030—growth at a pace of 29% annually starting in 2025.
BlackRock expects that standardized ‘claims’ on this computing capacity could be tokenized and transferred, used as collateral, and settled on-chain. The report notes that AI agents will ‘shop’ for the optimal computing resources in real time based on factors such as price, latency, and region. It assessed that an early example of this shift was shown through **Stripe**’s August acquisition of OpenRouter.
It also added an analysis suggesting that as traffic on public blockchain networks such as Ethereum (ETH) increases, demand for native tokens and the value of ‘blockspace’ could rise in tandem.
The authors describe AI as a “structural catalyst driving digital asset adoption,” pointing out that investors still underestimate the link between AI and crypto. At the same time, they drew a line, saying that agent-based payment activity and computing market liquidity remain limited for now.
Brian Armstrong’s AI-crypto thesis
This argument closely aligns with the story that crypto industry executives have been emphasizing for months. Coinbase CEO **Brian Armstrong** said in a post on X on July 26 that in the AI era, the importance of crypto is growing because agents can’t open bank accounts or wait for international transfers for days.
He emphasized that Coinbase’s x402 protocol and its own Layer 2 network, Base (Base), as well as USD Coin (USDC), already handle a significant portion of agent-based payments, and that BlackRock’s white paper also points to x402 as a potential emerging standard for ultra-fast machine-to-machine payments.
Next read: OpenAI Faces British Columbia Lawsuit Over 8-Victim School Shooting