Key content
BlackRock expects that standardized compute (operator resource) contracts will be circulated in the form of exchange-listed futures, with their capacity entitlements set as collateral and settled and cleared on programmable infrastructure (such as blockchain).
These projections are contained in a research report—“The Machine-Native Economy”—that BlackRock released last week.
Stripe’s acquisition of an open router, reported to be valued at over $7 billion in August, is presented as the first strategic signal of this shift.
BlackRock, the world’s largest asset manager, released a research report viewing computing power as “tradable financial assets” that can be used for futures, collateral, and on-chain settlement.
According to the report, computing power is being seen as “emerging as an independent large-scale economic resource, expanding into increasingly investable areas, and able to support a new type of digital asset.” Contract design for trading computing power is said to take the position that it is not “a question of possibility, but a matter of design.”
Though it doesn’t have a name, it is essentially a commodity market.
BlackRock directly compares computing power with traditional commodity markets. The report analyzes that “commodity resource markets have historically built their own trading infrastructure during their growth process, and as AI adoption expands, compute can follow a similar path.”
In particular, it expects that “once standardized products—including exchange-listed compute futures—appear, both compute providers and buyers will have access to more transparent price discovery and more efficient hedging tools.” Such standardized contracts create claims on compute capacity, and the report explains that these claims can be “expressed, transferred, have collateral settings applied, and settled through programmable infrastructure.”
BlackRock also acknowledges the barriers to overcome. Productivity differs by semiconductor generation, power cost differences by region are significant, and there is no standard for how to actually deliver contracted compute capacity and connect it all the way to cash settlement. However, it characterizes this as “important, but ultimately a problem that can be solved through design.”
Also to see: Bitget resumes withdrawals after a $388M hack…starting with Bitcoin
The $7B Stripe deal signals what’s coming.
The most concrete ongoing example presented by the report is Stripe’s agreement to acquire OpenRouter. OpenRouter is infrastructure that routes workloads across 400-plus AI models from more than 80 providers.
Major financial media outlets reported that the deal size is more than $7 billion as of August, but Stripe did not disclose the exact terms.
BlackRock views this acquisition as an “early strategic signal” that model routing and compute utilization optimization are becoming part of the financial infrastructure surrounding AI. In particular, the structure in which a payments company buys a compute routing firm is interpreted as a symbolic example of the “point where payments and compute meet” described in the report.
‘Inference’ sparks the market
The factor that changes the game is where AI systems actually use power.
Citing McKinsey’s estimate, the report says that by 2030, inference—not model training—will account for the largest share of data center power demand. It projects inference to take 43% of global data center power demand, while training will take 28%.
While demand for training is concentrated among a small number of large players, demand for inference is distributed across many companies and services. BlackRock views this structure as a typical set of conditions in which a “market” forms, not a “procurement contract.”
Based on the consensus outlook cited by the report, the combined revenues of the major hyperscale cloud segments are expected to reach about $1.1 trillion by 2030. This represents growth of 29% annually on average from 2025 levels.
Next to read: Solana ETF sees $188.10M inflow over 5 days…total net inflow surpasses $1.5B