On 19 August 2026 the Commodity Futures Trading Commission issued a request for comment on the listing of compute derivatives contracts, asking the public about compute cash markets, market oversight and manipulation risk, customer protection, and perpetual compute futures.1 CFTC 2026-08-19 The CFTC issued a request for comment on 19 August 2026 on the listing of compute derivatives contracts, covering compute cash markets, market oversight and manipulation, customer protection, and perpetual compute futures, with comments due 60 days after Federal Register publication. Open source The notice was published in the Federal Register on 21 August 2026, with comments due by 20 October 2026.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source The stake is not the paperwork: it is whether the cost of AI computing power becomes a hedgeable commodity price on a regulated US exchange, the way crude and power already are. We assess with moderate confidence that a listed compute contract is more likely than not to appear on a designated contract market within the next eighteen months, and with high confidence that the binding constraint will be the reference price, not the legal authority.

What the Commission actually asked

The request is not a proposed rule and creates no new obligations. It is a survey, and the shape of the survey is the news. The Commission grouped its questions into four areas: the characteristics of compute cash markets, including price transparency, liquidity, fungibility and standardisation; market oversight and manipulation risk, including settlement reliability and information sharing with compute providers; customer protection, including anti money laundering and disclosure questions; and whether perpetual contracts with no fixed expiry would serve the market better than dated futures.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source The same four headings appear in the Commission's own announcement.1 CFTC 2026-08-19 The CFTC issued a request for comment on 19 August 2026 on the listing of compute derivatives contracts, covering compute cash markets, market oversight and manipulation, customer protection, and perpetual compute futures, with comments due 60 days after Federal Register publication. Open source

Two of those four are the hard ones, and they are hard for the same reason. Under the Commodity Exchange Act, an exchange listing a contract must satisfy Core Principle 3, which requires that a settlement price be reliable, publicly available and timely and drawn from a liquid cash market, and Core Principle 4, which requires the exchange to run surveillance capable of detecting manipulation.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source A contract on compute has to point at a number. Today the underlying trade is mostly bilateral: negotiated capacity contracts between a small number of cloud operators and a small number of large buyers, priced privately. That is the gap the request is written around, and the Commission is asking the market to describe it rather than asserting it has been closed.

Why compute is being pulled into the futures perimeter now

The scale argument is doing the work. The request cites academic estimates putting annual gross compute service flow at roughly 430 billion to 1.3 trillion dollars, or about 1.4 to 4.0 percent of US gross domestic product.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source A cost base that large, moving that fast, with no standard forward curve, is a hedging vacuum. Chairman Michael Selig framed the move in competitive terms, saying the request is a first step toward clear rules for American compute markets, and tying a derivatives market for compute to the country's position in AI.3 Reuters 2026-08-19 Reuters reports the CFTC seeking comment on compute derivatives as AI demand grows, and quotes Chairman Michael Selig describing the request as a first step toward clear rules of the road for American compute markets and tying a compute derivatives market to US standing in AI. Open source That framing is an official statement of intent, not an independent assessment, and should be read as such.

Private activity had already moved first. Goldman Sachs and JPMorgan explored the space in June 2026, weighing futures referenced to graphics processing unit rental pricing, and Polymarket executed a first institutional block trade tied to AI compute infrastructure in the same month.4 PYMNTS 2026-08-19 Goldman Sachs and JPMorgan explored the compute derivatives market in June 2026, weighing futures tied to graphics processing unit rental pricing, and Polymarket executed a first institutional block trade linked to AI compute infrastructure in June 2026. Open source That sequence matters for reading the request: the Commission is responding to bank and platform interest that already exists, which raises the probability that comment letters will arrive with concrete contract designs attached rather than abstractions. The day after the request, the Commission convened the inaugural meeting of its Innovation Advisory Committee in Washington, chaired by Walt Lukken, with a session on artificial intelligence and computing in derivatives markets alongside crypto assets and prediction markets.5 CFTC 2026-08-20 The CFTC held the inaugural meeting of its Innovation Advisory Committee in Washington on 20 August 2026, chaired by Walt Lukken, covering crypto assets, artificial intelligence and computing in derivatives markets, and prediction markets. Open source We assess with moderate confidence that the two events were sequenced deliberately to put the compute question in front of exchange and platform executives while the comment window was opening.

Second order effects, and the ledger

The first order effect of a working compute futures market is that AI capital expenditure stops being a pure balance sheet risk and becomes partly a traded one. That redistributes advantage in specific ways.

Who gains. Exchanges and clearing houses gain first and most cleanly: a new contract class on an asset with the notional scale the request describes2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source is fee revenue and open interest they do not have to take from anyone. Neoclouds and mid sized GPU operators gain, because a public forward curve lets them presell capacity and finance builds against a hedged price instead of a negotiated promise. Banks that positioned early, including the two named as exploring GPU rental futures in June,4 PYMNTS 2026-08-19 Goldman Sachs and JPMorgan explored the compute derivatives market in June 2026, weighing futures tied to graphics processing unit rental pricing, and Polymarket executed a first institutional block trade linked to AI compute infrastructure in June 2026. Open source gain the market making franchise. Smaller AI developers gain in principle: a hedge is the substitute for the long term capacity contract they cannot get.

Who loses. The largest cloud providers lose the most, and the mechanism is price disclosure rather than regulation. Their advantage in bilateral capacity negotiation rests on the buyer not knowing what anyone else paid. A settlement price that is reliable, publicly available and timely, which is exactly what Core Principle 3 demands,2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source erodes that. Any index provider whose numbers become the reference also acquires an obligation it may not want, since the Commission is explicitly asking about surveillance and information sharing with compute providers.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source And retail sits in an uncomfortable place: the request raises customer protection and disclosure questions for what it treats as a geopolitically sensitive commodity,2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source which is a signal that retail access is contested rather than assumed.

The counter case

The strongest argument against the thesis is that compute is not fungible and may never be. A futures contract needs a deliverable or a reference that means the same thing to every participant. An hour on one accelerator generation, in one region, with one interconnect and one power contract, is not an hour on another. Electricity solved this with delivery nodes and hubs, but that took decades and a physical grid with metered flows. If commenters tell the Commission that no standardised unit exists, the honest outcome is a long silence rather than a rule, and the Commission has explicitly asked about fungibility and standardisation rather than assuming them.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source

There is a second failure mode. A request for comment is the cheapest action a regulator can take, and it creates no obligations of any kind.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source The Commission's public framing is ambitious,3 Reuters 2026-08-19 Reuters reports the CFTC seeking comment on compute derivatives as AI demand grows, and quotes Chairman Michael Selig describing the request as a first step toward clear rules of the road for American compute markets and tying a compute derivatives market to US standing in AI. Open source but ambition in a press statement is not a rulemaking calendar. For the thesis to fail it would be enough for the comment file to divide, for no exchange to self certify, and for the file to sit. We hold low confidence that any binding CFTC rule text on compute derivatives appears before the end of 2027.

What to watch

  • The comment file closes and who is in it. Comments are due 20 October 2026.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source If letters from the largest cloud operators are absent or argue that no reliable reference price exists, the settlement problem is unsolved and the timeline stretches.
  • A designated contract market self certifies a compute contract. Watch for the first exchange filing within roughly six months of the comment close. A listing that names a specific published index as its settlement source is the moment the reference price question is answered in practice rather than in theory.
  • A public compute price index acquires methodology governance. If an index provider publishes a surveillance and information sharing arrangement with compute providers during the next two quarters, that is the Core Principle 3 and 4 gap closing.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source
  • The perpetual question gets an answer. The request singles out perpetual compute futures.2 Baker Botts 2026-08 RFC issued 19 August 2026, published in the Federal Register 21 August 2026, comments due 20 October 2026; four categories of questions covering cash market transparency, liquidity, fungibility and standardisation, manipulation risk and information sharing with compute providers, AML and disclosure for a geopolitically sensitive commodity, and perpetual contracts; academic estimates of annual gross compute service flow of 430 billion to 1.3 trillion dollars, about 1.4 to 4.0 percent of US GDP; Core Principle 3 requires a reliable, publicly available and timely settlement price from a liquid cash market and Core Principle 4 requires surveillance; the RFC creates no new obligations. Open source If the first listed product is perpetual rather than dated, the Commission has accepted a contract design imported from crypto venues, and the retail access fight follows immediately.
  • The Innovation Advisory Committee returns to compute. The committee met for the first time on 20 August 2026 with compute on the agenda.5 CFTC 2026-08-20 The CFTC held the inaugural meeting of its Innovation Advisory Committee in Washington on 20 August 2026, chaired by Walt Lukken, covering crypto assets, artificial intelligence and computing in derivatives markets, and prediction markets. Open source A second session that produces a written recommendation, rather than a discussion, would be the clearest signal that the Commission intends to move from survey to rule.

The deeper point is that a regulator is now treating compute the way regulators treat oil, gas and power: as an input whose price needs a market, not just a contract. Whether that happens is a plumbing question about reference prices, and the answer arrives in the comment file, not in the rhetoric.