AI in action · Markets

Futures on Computing Power: The CFTC's Comment Period on Compute Derivatives Closes on Tuesday

The US Commodity Futures Trading Commission asked in August what rules a futures exchange should follow before listing contracts tied to AI computing power, covering cash markets, manipulation, customer protection and perpetual compute futures. Comments are due by October 20, 2026. Chairman Michael Selig called it "the first step toward establishing clear rules of the road for American compute markets".

The most AI-specific piece of US financial rulemaking this autumn is not about trading bots. It is about treating computing power as a commodity. On August 19, 2026, the Commodity Futures Trading Commission issued a request for comment on the listing of compute derivatives: futures and other contracts whose underlying is a unit of GPU or data-centre capacity, the input that AI developers now buy in bulk and whose price swings with every new model and every new chip. The request was published in the Federal Register on August 21 under RIN 3038-AF77, and the deadline for comments is Tuesday, October 20.

The CFTC's framing was unusually blunt for a regulator. "America cannot win the AI race without a robust derivatives market for compute," Chairman Michael Selig said in the accompanying statement, adding that the request was "the first step toward establishing clear rules of the road for American compute markets". He has elsewhere described compute as the most important commodity of the day. The request itself is narrower: it asks what factors a designated contract market should consider in meeting the Commodity Exchange Act's requirements when the underlier is compute, rather than wheat or crude oil.

The questions are the ones every new futures market has to answer. How deep and transparent is the cash market for compute, where prices are set in private contracts between cloud providers and their customers and vary by chip type, location and term? How would an exchange detect and deter manipulation when a handful of providers control most of the supply? What customer protections apply if speculators rather than hedgers dominate? And how should perpetual compute futures, contracts with no expiry modelled on the crypto market's perpetual swaps, be treated under rules written for dated contracts? The request does not propose a rule, approve a contract or authorise any exchange to begin trading.

The commercial case is straightforward. AI developers face a cost they cannot hedge, data-centre operators carry capacity they cannot sell forward at a known price, and the lenders financing both would like a reference price. A liquid compute futures curve would give all three what oil futures give airlines and refiners. The same curve would also give the market a daily, public price for the input behind every AI capital-spending plan, which is information the equity market currently infers from company guidance.

What happens after Tuesday is a proposed rule, or nothing, depending on what the comments say. The CFTC has been building toward this through its Innovation Task Force, set up in March to work with builders in crypto, AI and prediction markets, and through an advisory committee whose first meeting in August covered the same ground. For traders the practical effect is some way off; for anyone trying to value a company that buys or sells compute, the arrival of a traded price would change the arithmetic.

Sources: Reuters (via Yahoo Finance) — US CFTC seeks comment on compute derivatives as AI demand grows (Aug 19, 2026); Sidley Austin — CFTC Seeks Comment on Listing of Compute Derivatives Contracts; Baker Botts — CFTC Seeks Comment on Listing of Compute Derivatives, Signaling Forthcoming Market Rules