BlackRock’s new paper on the “machine-native economy” argues that autonomous AI agents will need stablecoins and tokenised compute. The thesis has gained significant attention in fintech, but on-chain activity presents a more cautious picture, while Europe’s rulebook adds another layer of complexity. The paper describes a shift from today’s Human → App → Bank → Merchant model toward an Agent → Agent → API → Settlement model.
BlackRock argues that AI agents buying data, calling APIs, and renting compute independently need payment methods that do not require human approval for every transaction. Stablecoins on public chains are presented as the only programmable, always-on settlement asset available at scale today, while tokenised bank deposits remain in trials. The paper also suggests that tokenised compute could become a tradable asset supporting financing and programmable settlement.
The rails for agent payments already exist. x402 , an open standard co-founded by Coinbase and Cloudflare, embeds payments into the HTTP “402 Payment Required” status code, allowing an API to charge an agent per call; the paper also references the Machine Payments Protocol. Usage remains limited, with reported activity suggesting that infrastructure is ahead of the demand it was built to serve.
For European builders, the machine-native economy runs into a specific rulebook. Under MiCA , euro stablecoins cannot pay interest to holders, while the ECB wants large issuers to rely less on bank deposits and its Pontes platform offers central bank money for wholesale tokenised settlement. Europe may develop a split model using public stablecoins for machine micropayments and central bank money for large-value transactions, while businesses test real agent services, spending controls, and tokenised deposits.