Regulator and Compute Futures: The New Raw Material

The thesis: compute is about to become a traded asset

AI compute capacity will become a liquid financial commodity within three years. This is a fact under construction, a trajectory readable in the activity of the U.S. regulator.

The consensus reads the request for public comment as a routine procedure. The consensus holds the wrong frame. This is a regime change, the birth act of a market that today remains invisible to the strategic models of most companies.

Confidence: high on the technology, medium on the timing. This article builds the thesis brick by brick.

What the regulator asked for, concretely

The Commodity Futures Trading Commission, the body that oversees futures markets in the United States, has opened a public consultation on contracts tied to AI compute power. The 60-day window opened on August 19, 2026, and two exchanges have already stated their intentions.

CME Group and Intercontinental Exchange intend to launch compute futures products as soon as the green light arrives. The two platforms dominate global derivatives. Their move signals conviction, money, and infrastructure ready to go. A regulator sees two competing candidates on the same product once that product stands finished by design, awaiting the stamp — rather than while it lingers as a hypothesis.

A future is an agreement to buy or sell something later at a price fixed today. It serves to lock in costs ahead of time. Data centers would use the same mechanism that farmers apply to harvests and airlines apply to fuel. The precedent matters: each time a resource becomes central to production, sooner or later it acquires a derivatives market that stabilizes its price. AI compute today sits as central as fuel was to twentieth-century aviation.

Why the consensus misreads this move

90% of analysts get the present right. They get the pace of change wrong. They see a technical consultation, they file the topic among regulatory curiosities.

The figure everyone watches is the slowness of regulatory processes. The figure that truly predicts differs: when two competing exchanges announce the same product ahead of approval, the product arrives. The regulatory question becomes a matter of design and calendar.

Here the strongest counter-argument deserves acknowledgment. Compute is hard to standardize. A token, a GPU-hour, and an inference cycle stand as heterogeneous units, far from the homogeneous barrel of oil. Those who read the consultation as a distant exercise grip precisely this: absent a reference standard, a future lacks a measurable underlying.

The objection holds real weight, and it shifts the date while leaving the direction intact. This is the true challenge the regulator must solve, and the reason the launch slips by quarters, staying inside the horizon of a few years. Exchanges have solved more complex standardization problems: they built electricity and bandwidth indices starting from equally fragmented underlyings. Compute follows the same path.

The cost curve says who is right

The commoditization of an asset follows a cost curve. Solar offers the map: the International Energy Agency documented a drop in photovoltaic prices close to 90% across the 2010-2020 decade, with high values at the start of the period, halved at the midpoint, and marginal at the end.

Compute follows the same shape. The price per unit of performance of foundation models has collapsed in a comparable way. According to my analysis, GPT-4-equivalent performance travels toward 0.001 dollars per token by the end of 2026, along a descent that starts from values tens of times higher two years earlier.

A limit belongs on the evidence. The solar curve measures the production of a physical resource; the compute curve measures the price per performance, distinct from the price of raw hardware. These are two different quantities. Still, the point for the futures market holds: what a derivative requires is a price that moves predictably enough to be hedged, rather than a price that falls.

A commodity with a price in vertical free fall and demand in exponential growth generates volatility. Volatility creates the need for hedging. Hedging creates the futures market. This is the causal chain, and it is the engine the exchanges have already read.

Cliff event: when adoption jumps

Adoption of a new financial instrument grows slowly, then jumps. The jump arrives when the first liquid contract attracts the second tier of participants.

Cliff event: the first compute future listed and traded on CME or ICE, by the end of 2027, with volumes crossing the threshold of useful liquidity. At that date compute stops being an operating cost line and becomes a financial position in full.

The mechanism is clear. Speculators bring liquidity, liquidity makes the contract usable, usage attracts data centers as natural hedgers. The cycle feeds itself once it clears critical mass. The consequence for anyone leading a company is concrete: the cost of compute stops being a private negotiation and becomes a public number, readable every day. Those who read it plan better. Those who ignore it pay the spread.

Three categories that will change shape

Three sectors will operate differently by 2028, in forms the market today still misprices:

Cloud providers and data centers

They stop negotiating capacity in opaque bilateral contracts. They lock the price on the market, as an airline does with jet fuel. The margin shifts toward whoever manages the position best, toward whoever reads the curve.

Venture capital and growth equity

They reassess the cost of compute as a hedged variable. An AI startup with predictable compute costs presents a different risk profile. Deal size and covenants will change accordingly, and funds that ignore the hedge will overprice the risk.

Technology procurement

It faces the most delicate choice. Locking a vendor today onto a fixed long-term rate means betting against a market about to give a public price to the same resource. The wrong multi-year contract becomes a losing short position.

My position, and what would overturn it

The position is sharp: compute is the next great financial commodity, and the regulator's move marks act one of a structural transformation.

This ties to a core thesis of mine. Foundation models will commoditize within 18-24 months, and value will migrate to the application layer. An asset that commoditizes acquires a derivatives market. Compute follows the same script.

What would change my analysis? A sharp signal: the withdrawal of CME and ICE from their plans, or a regulator decision that closes the door on compute contracts. That fact would zero out the thesis.

The forecast

Explicit forecast: at least one major exchange between CME Group and Intercontinental Exchange will list an AI compute futures contract, tradable live, by December 31, 2027.

Confidence: 70%. Horizon: end of 2027. The causal mechanism is the convergence of hedging demand, price volatility, and exchange infrastructure already in place.

Kill signal: at the stated date, zero compute futures contracts stand listed and traded on a CFTC-regulated exchange. That figure, verifiable, would close the case.