Table of Contents
BlackRock's Digital Assets Research team published a whitepaper last week arguing that autonomous AI agents could become an overlooked source of demand for stablecoins and blockchain infrastructure. The paper is titled The Machine-Native Economy: How digital assets connect intelligence, commerce, and compute. Its authors are Will Su, Robert Mitchnick, Jay Jacobs, and William Helm.
The argument is that agentic AI systems, which plan and carry out multistep tasks with limited human oversight, will need to pay for data, services, and compute without a human approving each transaction. BlackRock's position is that conventional payment rails are poorly suited to that pattern and stablecoins are better fitted to it.
The firm is careful about what it is claiming. BlackRock presents these as projections rather than established outcomes and states plainly that the ecosystem remains nascent.
BlackRock frames the convergence in one line: "AI represents machine-native intelligence, while digital assets represent machine-native money."
Key Takeaways
- BlackRock published The Machine-Native Economy the week of September 22.
- It argues AI agents will need always-on, sub-cent payment rails.
- Stablecoins are named as the instrument most likely to lead transactional use.
- The paper floats tokenized claims on computing capacity as a new asset class.
- BlackRock stresses that agentic payment activity remains limited today.
What the Paper Argues
BlackRock identifies three areas where AI and digital assets overlap: tokenization, agent-driven payments, and a future market for computing power. The payments case rests mostly on stablecoins.
The reasoning is about transaction shape rather than ideology. ACH and card networks carry onboarding requirements and settlement economics that make them awkward for very low-value, always-on transfers, while crypto rails handle high-frequency sub-cent machine-to-machine payments such as API calls and on-demand compute.
Stablecoins get the lead role because a stable unit of account gives predictable pricing and settlement. Native cryptoassets and tokenized real-world assets are described as complementary programmable instruments.
The Numbers Behind the Claim
BlackRock puts circulating stablecoin market capitalisation above $300 billion as of September 2026. It cites adjusted stablecoin transaction volume above $11 trillion in 2025, which places it in the same range as Visa and Mastercard annual volumes.
The growth comparison is the more interesting figure. Stablecoin volume has compounded at roughly 80% a year since 2020, against about 8.5% for ACH, though ACH still moved $93 trillion in 2025.
Those totals are dominated by trading and transfers rather than payments for goods and services. The WTO put stablecoins at about 3% of international payments this month, which we covered in our report on the WTO stablecoin report.
Tokenized Compute Is the Larger Claim
The part of the paper that goes furthest is not about payments at all. BlackRock suggests standardized claims on compute capacity could be tokenized, transferred, pledged as collateral, and settled through programmable infrastructure.
In that model, agents compare computing capacity on price, performance, latency, location, and hardware specialization, then provision resources and settle payments automatically. That would make compute a tradeable financial asset rather than a procurement line.
It is also the least developed idea in the paper. BlackRock notes that compute-market liquidity is still limited, which is a significant caveat for an asset class that would need depth to function.
The Protocols and Rails Named
BlackRock points to protocols including MCP, A2A, and x402 for data access, agent coordination, and machine payments. Coinbase's x402 is highlighted as an early working example, built around an unused HTTP response code that lets software pay for an API call within the same request that asks for it, with no account setup.
Other agent payment efforts named in coverage include ACP and Stripe's machine payments work. The institutional settlement venues are being built in parallel, including the chain Circle opened this month with BlackRock itself among the validators, which we covered in our report on the Arc mainnet validators.
The Agents Have Not Arrived
The strongest counterweight to the thesis comes from usage data. Blockchain intelligence firm TRM Labs found that AI agents currently account for as little as 0.6% to 7.5% of payment volume on x402.
That range matters because x402 is the protocol most often cited as evidence the machine economy is underway. Most of the activity on it is not agentic.
BlackRock does not claim otherwise. The paper describes a forecast about customers who have not yet shown up, which is a reasonable thing for a research note to do and a poor basis for a market call.
What Could Undercut the Thesis
The clearest challenger is tokenized bank deposits. If banks bring programmable deposits to production, the argument that stablecoins are the only always-on programmable settlement asset weakens considerably.
Regulation is the second variable. The money agents would transact in is being brought inside bank-style supervision, with the Federal Reserve proposing reserve, capital, and custody rules this month, which we covered in our report on the Fed reserve proposals. Compliance requirements designed for human-initiated payments may not map cleanly onto sub-cent machine transfers.
FAQs:
1. What is BlackRock's Machine-Native Economy paper?
A research whitepaper from BlackRock's Digital Assets Research team, published the week of September 22, 2026, titled The Machine-Native Economy: How digital assets connect intelligence, commerce, and compute. It argues that autonomous AI agents paying for data, services, and compute could become a significant and underappreciated source of demand for digital assets.
2. Why does BlackRock favour stablecoins for AI payments?
Because a stable unit of account gives predictable pricing and settlement, and blockchain networks settle around the clock without requiring a bank account. The paper argues ACH and card networks carry onboarding requirements and settlement economics that make them less suited to always-on, sub-cent machine-to-machine transactions.
3. What is tokenized compute?
BlackRock's proposal that standardized claims on computing capacity could be tokenized, transferred, pledged as collateral, and settled programmatically. Agents would compare capacity on price, performance, latency, location, and hardware specialization, then provision and pay automatically. The paper notes compute-market liquidity remains limited.
4. Are AI agents actually making stablecoin payments today?
Very few. TRM Labs found AI agents account for as little as 0.6% to 7.5% of payment volume on Coinbase's x402 protocol, which is the infrastructure most often cited as evidence of machine commerce. BlackRock itself says the ecosystem is nascent and presents its case as projection rather than outcome.
5. What numbers does the paper cite?
Circulating stablecoin market capitalisation above $300 billion as of September 2026, and adjusted transaction volume above $11 trillion in 2025, comparable to Visa and Mastercard annual volumes. It notes stablecoin volume has grown roughly 80% annually since 2020, against about 8.5% for ACH, which moved $93 trillion in 2025.
Disclaimer:
This content is provided for informational and educational purposes only and does not constitute financial, investment, legal, or tax advice; no material herein should be interpreted as a recommendation, endorsement, or solicitation to buy or sell any financial instrument, and readers should conduct their own independent research or consult a qualified professional.
