Invest2 publishers3 min readPublished
TRM sizes this year's AI agent payments on x402 at $4 million at the outside
BlackRock's new paper argues stablecoins are the rail built for sub-cent machine payments, and TRM Labs, examining $52.7 million of Coinbase x402 settlements, could attribute between 0.6% and 7.5% of the value to AI agents.
The Investor · Invest desk
What happened
- BlackRock published a paper this week called The Machine-Native Economy, arguing that AI, not regulation or institutional buying, could be one of the biggest overlooked drivers of crypto demand.
- Its core claim is that stablecoins, not banks or card networks, can handle the sub-cent round-the-clock payments AI agents will need to make on their own.
- The paper also floats tokenized claims on computing power as an asset class, standardized like oil or wheat futures, against roughly $1.1 trillion of projected 2030 cloud revenue at three vendors.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint The only public measurement of agent payment volume is too small to size a market from, so any allocation made on this thesis in 2026 is a bet on infrastructure the paper itself says precedes the demand.
- decision A team metering an API in fractions of a cent now chooses between building account-based billing and accepting stablecoin settlement, because the rails shipped ahead of the buyers.
- exposure A 12.5x gap between the floor and ceiling of the only public estimate means anyone underwriting agent payment growth is underwriting a definition of "agent", not a measurement of one.
- capability Standardized compute claims would make cloud capacity financeable before it is consumed. Lenders can write those deals without an agent paying for anything.
Take TRM Labs' ceiling of 7.5% on the $52.7 million of x402 settlements it examined [7] and AI agents paid for $3.95 million of things this year [17]. Take the 0.6% floor and the figure is $316,000 [17]. The ceiling is 12.5 times the floor [20], and the firm's stated reason for the spread is that most of the traffic looked more like ordinary automated scripts than genuine autonomous agents [8]. Against the $11 trillion of adjusted stablecoin volume BlackRock counts for 2025 [9], the upper figure is one part in 2.8 million [18].
Whatever that number is, BlackRock's case rests on a growth differential. Stablecoin volume at $11 trillion is 11.8% of the $93 trillion that moved over ACH in 2025 [19], and stablecoin volume has compounded at roughly 80% a year since 2020 against about 8.5% for ACH [11]. Hold both rates and the smaller number passes the larger one in a little over four years [21]. I doubt either rate survives four years, because 80% is what a base does while it is still small, but that extrapolation is the demand case.
The second idea in the paper is an asset class that does not exist yet. "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 said [5], with the paper comparing such contracts to oil or wheat futures [6]. The revenue base it points to is roughly $1.1 trillion of combined 2030 cloud revenue at Amazon, Microsoft and Google, on analyst estimates BlackRock cites [15]. Such a market runs on lenders and forward prices, and nothing in the paper requires agent payments to reach scale first [16].
The settlement layer is already built. x402, Coinbase's protocol built on the old HTTP 402 code, lets software pay for a data feed or an API call in the same request that asks for it, with no account and no human sign-off [12]. Amazon wired stablecoin payments into its AI cloud tools with Coinbase and Stripe, so agents can pay for APIs, data feeds and bookings mid-task [13]. Google extended its Agent2Agent framework to cards, stablecoins and real-time bank transfers, backed by Coinbase and the Ethereum Foundation [14]. A team pricing a metered API in fractions of a cent has a rail for it now, and the same rail carries the payment whether the payer is an agent or a scheduled job.
The paper is explicit that it is not claiming this happens at scale soon [16], so the thing to falsify is the growth rate, not the agent count. If 2026 stablecoin volume grows at something like ACH's 8.5% while x402 settlements climb on script traffic, then stablecoins are automated billing for customers who already had payment methods, and the four-year crossover never arrives. The other way this goes wrong for BlackRock is narrower and more likely: compute claims get written, but by the cloud vendors and their lenders in ordinary contracts, with nothing tokenized and no stablecoin in the chain. Either way, the $11 trillion is the number to test.
What to watch
- Whether Amazon's or Google's agent payment layers publish settlement volumes, which would give a denominator TRM's sampling of x402 cannot.
- Whether any venue lists a standardized compute contract of the kind the paper describes, and who takes the other side of it.
- Whether BlackRock follows the research with a product, and which desk owns it: Digital Assets Research or iShares.