BlackRock’s machine economy: stablecoins, tokenized compute and the gap between vision and usage

BlackRock’s The Machine-Native Economy presents a possible next step for digital assets: autonomous AI agents could pay for data, software access and computing resources, while tokenized claims on compute capacity could be traded or financed. Its central idea is a forward-looking thesis, not proof that agent commerce is already a large market. 410

Key takeaways

• Stablecoins are the proposed payment tool. BlackRock argues that machine-to-machine purchases may need programmable payment infrastructure; coverage identifies Ethereum and Circle’s Arc as potential rails, and Coinbase’s x402 as a payment protocol. 710
• The headline stablecoin figures aren’t agent-payment figures. The report points to more than $300 billion in stablecoins outstanding and over $11 trillion in adjusted stablecoin transaction volume in 2025. Those totals reflect stablecoin activity broadly. 212
• Tokenized compute is a potential market, not an established asset class. A projection of roughly $1.1 trillion in combined annual cloud revenue by 2030 describes the scale of the underlying market; it is not a forecast of tokenized-compute sales. 213
• Current agent-payment estimates are far smaller than the vision. TRM Labs found substantial x402 settlement activity, but estimated that only a small share of screened commerce appeared agentic. 17

Why BlackRock sees stablecoins as a fit for AI agents

An agent that independently requests data, calls software services or rents compute may need to pay without a person stepping in to complete each transaction. BlackRock’s argument is that programmable networks and digital assets could support that kind of machine-to-machine payment and settlement. The report describes stablecoins as one possible instrument on those rails; it does not establish that agents already account for a significant share of stablecoin use. 10

Coverage of the report names Ethereum and Circle’s Arc as settlement venues and Coinbase’s x402 as an emerging protocol for machine-initiated payments. 7 These are part of the proposed infrastructure, not evidence that autonomous agent commerce has scaled across those networks.

The market figures give context, but need careful interpretation. BlackRock’s cited stablecoin totals—more than $300 billion outstanding and over $11 trillion in adjusted transaction volume in 2025—cover stablecoin activity overall. They cannot be read as payment volume generated by AI agents. 212

How tokenized compute could become an asset

The report also considers whether claims on computing capacity could be represented as digital tokens and potentially traded, financed or used as collateral. That would make compute capacity a possible digital-asset use case, but the available reporting describes it as an opportunity under discussion—not a mature market with established trading activity. 1334

The roughly $1.1 trillion figure cited for combined cloud revenue by 2030 helps illustrate the potential size of the underlying compute economy. It is a projection for cloud businesses, not a prediction that tokenized compute will generate that amount of revenue. 213

Reactions and examples: what the evidence supports

Raoul Pal’s public post broadly agreed with BlackRock’s focus on stablecoins and tokenized compute, while arguing that identity, contracts and other resources could also become part of a machine economy. That is Pal’s wider interpretation, not evidence that those additional markets already exist. 41

Separately, a report citing XRPL AI Hub data says agents using the XRP Ledger have increasingly settled with regulated stablecoins rather than XRP. That offers a reported example consistent with the broader thesis, but the available coverage does not provide enough detail to quantify the shift or establish how representative it is. 33

The available sources document x402 as a Coinbase-led protocol, but do not substantiate a specific public response from Brian Armstrong to this report. 7

The reality check: x402 activity is not the same as agent spending

TRM Labs identified about $52.7 million across 198.9 million settlement transactions handled by known x402 facilitators on Base, Solana and Polygon since May 2025. But that total measures facilitator-mediated protocol activity, not confirmed purchases by autonomous AI agents. 17

When TRM assessed screened x402 commerce, it estimated that 0.6%–7.5% appeared agentic. A separate report puts the resulting monthly AI-agent payment run rate at roughly $5,000–$11,000. That monthly figure is an estimate with a narrower scope, not a definitive measure of every AI agent’s spending. 117

The distinction matters: a large transaction count or settlement total can show that payment infrastructure is being used without proving that autonomous agents are driving most of the activity. BlackRock’s proposal may describe a future source of demand, but the evidence cited here suggests that verified agent commerce remains nascent. 117