BlackRock says AI agents could become crypto’s next source of demand

BlackRock’s digital assets research team has published a whitepaper arguing that autonomous AI agents, not traders or treasuries, could be the next structural source of demand for stablecoins and blockchain settlement. The data on actual agent payments today is far thinner than the thesis, and the paper says so itself.

• BlackRock’s “The Machine-Native Economy,” published in September 2026, frames AI as machine-native intelligence and digital assets as machine-native money.
• The paper points to Coinbase’s x402 protocol and stablecoins as the payment layer for high-frequency, low-value machine-to-machine transactions, and to tokenized compute as a possible new asset class.
• Real usage is still small: TRM Labs found that only 0.6% to 7.5% of $52.7 million in x402 settlements it examined likely came from AI agents.

• $300B+ stablecoin market cap as of September 2026, per RWA.xyz data cited by BlackRock
• $11T adjusted stablecoin transaction volume in 2025, per Visa/Allium data cited in the paper
• $1.1T projected combined 2030 cloud revenue for AWS, Azure and Google Cloud

The world’s largest asset manager is making the case that crypto’s next demand wave will come from software rather than people. In “The Machine-Native Economy: How Digital Assets Connect Intelligence, Commerce, and Compute”, BlackRock head of digital assets research Will Su and head of digital assets Robert Mitchnick, writing with iShares colleagues Jay Jacobs and William Helm, argue that agentic AI systems that can act and pay on their own need a kind of money that card networks and bank rails were not built to provide. The paper was first reported by Decrypt.

BlackRock pitches stablecoins as the payment rail agents need

The core argument is about fit. An AI agent that buys a data feed, pays for an API call or rents GPU time may need to make thousands of tiny payments a day, around the clock, with no human approving each one. BlackRock notes that existing rails such as ACH and cards already support heavy automation, but says their onboarding requirements and settlement economics make them a poor match for always-on, very low-value transactions.

Stablecoins are the paper’s answer. BlackRock cites RWA.xyz data showing more than $300 billion in circulating stablecoins in September 2026, and Visa/Allium figures showing $11 trillion in adjusted stablecoin transaction volume in 2025. That volume grew at roughly 80% a year from 2020 to 2025, compared with about 8.5% a year for ACH, according to the paper. BlackRock adds a caveat worth keeping: the adjusted stablecoin figures filter out exchange flows, internal transfers and bot activity, so they are not directly comparable to traditional payment data. ACH still processed about $93 trillion in 2025.

Coinbase’s x402 sits at the center of the machine-payments thesis

The paper leans on x402, the open protocol developed by Coinbase that revives the web’s long-dormant HTTP 402 “Payment Required” status code. With x402, a server can ask for payment inside a normal web request, and a piece of software can pay in stablecoins, usually USDC, and receive the data or service immediately. No account, no card on file, no human in the loop.

BlackRock also points to big tech adoption. Amazon has added stablecoin payments to its AI cloud tooling through Coinbase and Stripe, and Google’s Agent2Agent framework supports cards, stablecoins and real-time bank transfers, with backing from Coinbase and the Ethereum Foundation. The x402 ecosystem has been widening on the crypto side too, with NEAR joining the x402 Foundation this week and Solana and Google launching Pay.sh earlier this year.

For native tokens, the paper is careful. It says activity on permissionless networks could drive demand for blockspace and validator services, but whether any of that value reaches a network’s token depends on its fee, staking and gas-sponsorship design. It names Ethereum and Circle’s Arc, where USDC is the native gas asset, as examples of the two ends of that spectrum.

Tokenized compute could be the bigger long-term market

The paper’s most ambitious section treats compute itself as a future digital asset. Sell-side consensus compiled by Bloomberg puts combined 2030 revenue for AWS, Microsoft’s Intelligent Cloud and Google Cloud at $1.1 trillion, and BlackRock cites Goldman Sachs estimates of more than $5 trillion in cumulative AI capital spending from 2025 to 2030.

“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 Digital Assets Research, The Machine-Native Economy

The idea is that agents could source, price and pay for compute directly, settled on-chain against standardized, tokenized claims on capacity. BlackRock flags real obstacles: chips differ by generation, energy costs vary by region, and there is no agreed standard yet for delivering or cash-settling capacity. It suggests commodity market tools such as basis markets and contracts for difference as a template.

On-chain data shows agent payments are still a sliver of activity

The gap between thesis and usage is wide. Blockchain analytics firm TRM Labs examined $52.7 million in x402 settlements in 2026 and estimated that only 0.6% to 7.5% of that traffic likely originated from AI agents. Most x402 volume today is still people and conventional software, not autonomous agents.

BlackRock does not hide this. The paper calls the ecosystem nascent, says agentic payment activity and compute-market liquidity remain limited, and describes machine-native payments as an area worth monitoring rather than a forecast. It also carries the standard disclaimer that it is not investment advice.

peaq brings machine identities onto Solana

The agent economy BlackRock describes is mostly software paying software. peaq is building the physical side of it, and this month it moved that work onto one of crypto’s most used chains. peaq launched its Economics 2.0 model on September 7, and the next day Wormhole Labs’ Sunrise gateway listed PEAQ on Solana as the canonical version of the token, according to Crypto Briefing. peaqOS then went live on Solana on September 16, letting robots and IoT devices activate directly on Solana, receive a verified on-chain identity and bond PEAQ.

Machine peaqID documents and NFTs now live on Solana, while subscription bonds stay on the peaq chain. Each machine builds a Machine Credit Rating from verified on-chain activity, which can later support financing. peaq says its network spans more than 6 million machines across 60-plus applications, with about 3.3 million positioned to bond PEAQ through the Sunrise integration and just over 240,000 activated by early September. Doosan Robotics deployed its first A-Series cobot on the system by September 17. The design is also a live test of the point BlackRock makes about token value: operators lock PEAQ as collateral to activate a machine, so demand for the token is tied to machines actually operating. It builds on peaq’s earlier move to bring World ID to robots.

“We’re still very early in the adoption of stablecoin payments by AI agents.”

Ashton Addison, CEO of Crypto Coin Show

Addison expects that to change quickly. “As the machine economy picks up over the next 12 months, we should see these numbers start growing exponentially,” he said. “With machine IDs moving onto Solana, the machine economy is reaching more mainstream blockchains, and that makes stablecoin payments much easier to adopt.”

The CCS read. The paper does not move markets on its own, but who wrote it matters. When BlackRock’s digital assets team, the same group behind the BUIDL tokenized Treasury fund, puts AI agents next to stablecoins and tokenized assets as a demand driver, it gives institutional allocators a framework for owning crypto that has nothing to do with speculation. The honest read is that this is an early, infrastructure-first story. The rails are being built by Coinbase, Google, Amazon and Stripe, while real agent volume is still measured in single-digit percentages of a $52.7 million sample. The value will likely accrue first to stablecoin issuers and the networks that settle their transfers, and only later, if at all, to tokens whose fee design captures that activity.

Watch the share of x402 and similar settlements that analytics firms can attribute to agents over the next few quarters. If that 0.6% to 7.5% range climbs meaningfully, BlackRock’s thesis moves from framework to measurable demand. Also watch for the first standardized compute contracts or tokenized capacity products from a major cloud or exchange partner.