In brief

• BlackRock's new "Machine-Native Economy" paper argues stablecoins are better suited than banks or card networks to handle the sub-cent, 24/7 payments AI agents will need to make on their own.
• The report floats a new asset class: tokenized claims on computing power.
• A separate study from blockchain intelligence firm TRM Labs found AI agents currently account for as little as 0.6% to 7.5% of payment volume on Coinbase's x402 protocol

BlackRock thinks the robots are going to need a bank account.

The world's largest asset manager published a research paper this week arguing that artificial intelligence, not new regulation or fresh institutional buying, could become one of the biggest and most overlooked drivers of demand for crypto. Per the firm, AI agents are about to start paying for things themselves, and stablecoins are the instrument best built for that job.

“As agents become more capable and persistent, standardized claims on compute capacity could become a significant digital asset use case for financing and programmable settlement,” Blackrock says.

The paper, titled "The Machine-Native Economy," comes from BlackRock's Digital Assets Research team, led by Will Su and Robert Mitchnick, alongside the firm's U.S. equity ETF and iShares product leads.

Its core idea is a tidy bit of wordplay. “AI represents machine-native intelligence, while digital assets represent machine-native money,” BlackRock says, arguing the two were always headed toward each other.

The paper leans on "agentic AI"—AI systems that can plan and carry out multi-step tasks on their own, calling outside tools and services with limited human hand-holding. Agents are AI that don’t just recommend a flight, but book it, pay for it, and hand you the receipt.

That kind of autonomy runs into a problem fast: someone, or something, has to pay the bill.

Traditional payment rails weren't built for that. Opening a bank account or a credit card requires a human with an ID, and card-network fees make sense for a $40 purchase but not a fraction-of-a-cent API call.

BlackRock's answer is stablecoins—cryptocurrencies pegged to a stable asset like the U.S. dollar, designed to hold a steady price instead of swinging like Bitcoin. They can move around the clock, settle almost instantly, and don't need a bank teller.

Adjusted stablecoin transaction volume topped $11 trillion in 2025, per the report, landing in the same range as Visa and Mastercard's annual payment volumes. It's still dwarfed by the $93 trillion that moved over the traditional ACH bank transfer system that year, but stablecoin volume has grown roughly 80% a year since 2020, compared with about 8.5% for ACH.

“The rise of agentic AI and machine-tomachine payments will likely increase demand for blockchains and other programmable payment infrastructure; stablecoins, native cryptoassets, and other on-chain assets can serve as machine-native instruments for payment and settlement across these rail,” the paper reads.

The plumbing already exists

Some of this machinery is already live. x402, a protocol built by Coinbase around the old, mostly unused "HTTP 402: Payment Required" web code, lets a piece of software pay for a data feed or an API call in the same request that asks for it with no account, no human sign-off required.

Amazon has already wired stablecoin payments into its AI cloud tools with Coinbase and Stripe, letting agents automatically pay for APIs, data feeds, and bookings mid-task. Google built its own agent-payments layer with backing from Coinbase and the Ethereum Foundation, extending its Agent2Agent framework to handle cards, stablecoins, and real-time bank transfers.

But real-world usage is still thin. Blockchain analytics firm TRM Labs examined $52.7 million in x402 settlements this year and found AI agents likely accounted for somewhere between 0.6% and 7.5% of that value—most of the traffic looked more like ordinary automated scripts than genuine autonomous agents.

BlackRock isn't claiming this is happening at scale soon. Instead, it's betting the plumbing is being built ahead of the moment it does.

Running AI is expensive, and demand for cloud computing capacity keeps climbing. BlackRock cites analyst estimates putting combined 2030 revenue for Amazon, Microsoft, and Google's cloud divisions at roughly $1.1 trillion.

The paper envisions "compute" packaged into standardized contracts, similar to how oil or wheat futures work today, that could be bought, sold, used as collateral, or settled automatically on a blockchain. An AI agent, in this scenario, could shop around for the cheapest available server capacity and pay for it directly, job by job, without a person negotiating a cloud contract.

BlackRock's own figures put stablecoins' circulating market cap above $300 billion as of September 2026, with more than $11 trillion in adjusted transaction volume moved in 2025 alone. The firm's bet is that AI agents, not human spenders, drive the next leg of that growth as autonomous systems begin paying their own bills.