Part of the AI Regulation News hub.
The CFTC is asking whether an exchange can lawfully list a futures contract on AI compute, and says compute may not yet be fungible, standardised or liquid enough to support one
This is the first time a US financial regulator has asked, on the record, what it would take to trade the cost of running a large language model. The Commission's own preliminary answer is unflattering: the cash market is opaque, bilateral and concentrated, and the firms best placed to move a settlement index are the same firms that sell the capacity.
Bottom line: A request for comment, not a proposed rule. It creates no obligation and changes no regulation. What it does is put the Commission's preliminary views about compute markets into the record, and those views set the bar a designated contract market will have to clear before it can list a compute contract.
Who this affects: Designated contract markets and swap execution facilities contemplating compute products, futures commission merchants and introducing brokers, compute and cloud capacity providers whose posted rates could feed a settlement index, price reporting agencies and index administrators, AI companies with material compute cost exposure, and the counsel and risk functions advising any of them.
Issue date: Published 21 August 2026 at 91 FR 54259. Comments must be received on or before 20 October 2026.
What changed: Nothing yet, legally. The change is informational: the Commission has stated preliminarily that compute may not yet exhibit fungibility, standardisation and sufficient liquidity, and has asked whether a contract may settle to a price the Commission cannot observe, verify or surveil.
Analysis: Read question 2(b). It describes, without naming anyone, a structure in which a capacity provider administers the posted rates that feed an index, while the remaining transactions execute on venues a small number of participants dominate. The Commission is asking what would stop that provider from moving the index by adjusting a rate or steering capacity during the observation window. That question is the centre of gravity of the whole document.
Primary sources: Request for Comment, official GPO text (91 FR 54259) · Same document, GPO PDF
- Instrument
- Request for Comment on the Listing of Compute Derivatives Contracts
- Authority
- Commodity Futures Trading Commission
- Citation
- 91 FR 54259, 21 August 2026, pages 54259 to 54264
- Regulatory identifier
- RIN 3038-AF77; 17 CFR Parts 1 and 38
- Jurisdiction
- United States, federal
- Status
- Request for comment. Open.
- Bindingness
- None. A request for comment imposes no duty and amends no rule. Its weight is evidentiary: it records the Commission's preliminary understanding of compute markets.
- Issue date / next deadline
- Issued 21 August 2026. Comments must be received on or before 20 October 2026.
- Review status
- A significant regulatory action under section 3(f) of Executive Order 12866, reviewed by the Office of Management and Budget.
- Editorial Note
- Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.
- Primary source
- https://www.govinfo.gov/content/pkg/FR-2026-08-21/html/2026-17163.htm
What is actually being proposed for trading
The underlier is access to computing power. The Commission's own definition is worth reading closely, because it ties the product to the AI stack rather than to hardware generally: compute is the processing power primarily used by the large language models at the centre of the artificial intelligence economy.
In practice, the Commission preliminarily understands, the commodity underlying a compute futures contract would typically be access to rented compute capacity from hardware the purchaser does not own. Its worked example is the hourly rental price of compute from a B200. It then flags an alternative underlier that would change the product substantially: access to a stated volume of LLM inference tokens.
The scope is not limited to futures on an exchange. A footnote extends the request to swap execution facilities listing contracts that settle to the price of compute, and to physically settled compute swaps. Perpetual compute futures get their own block of questions at the end.
The Commission's stated purpose is standardisation. It believes responses from industry may enable it to advance the standardisation of such products in a manner that promotes transparency, liquidity, and responsible innovation.
The Commission's preliminary view is the news
A request for comment usually reveals little. This one states a position, and the position is sceptical.
The Commission preliminarily understands that compute markets are fragmented and that price formation primarily occurs in opaque bilateral transactions, hindering the availability of current and historical price data, with consensus still forming about what the underlying compute commodity even is. It adds that dominant market participants may wield significant pricing power that may lead to manipulability, preferential pricing arrangements and unfair market dynamics, and that pricing varies dramatically across providers, regions and contract structures.
From that it draws the conclusion that matters to anyone drafting a contract specification: compute may not yet exhibit certain of the characteristics of commodities that typically underlie a commodity derivatives market, including fungibility, standardisation, and sufficient liquidity.
Those three words are the ones to plan against. Fungibility, standardisation and liquidity are not rhetorical. They map onto the statutory Core Principles a designated contract market must satisfy, and the Commission has now put on the record that it is not yet satisfied they are present.
The manipulation problem, stated plainly
The second block of questions is where the document stops being exploratory. Core Principle 3 requires a designated contract market to list only contracts not readily susceptible to manipulation. The Commission asks what a compute contract that cash settles to an index calculated over predominantly bilateral and privately priced transactions would have to demonstrate to satisfy that standard.
Then it describes the structural conflict directly. Certain published compute price series, it notes, are constructed in whole or in part from posted or listed rates that the compute capacity providers themselves administer, with the remainder of transactions executed on venues that a small number of participants operate or dominate. It asks what protections would prevent a capacity provider from manipulating a cash settlement index by adjusting a posted rate, by directing capacity onto or away from a venue whose transactions feed the index, or by executing or declining to execute transactions during the observation window.
A related question asks whether the parties best positioned to influence the reference price are the same parties that supply capacity, and whether that differs from other commodities. The Commission does not answer it. Anyone responding on behalf of an index administrator should expect that question to be the one their submission is judged on.
Core Principle 4 gets its own question: what surveillance capabilities would be necessary, and are those capabilities presently feasible from a technological, operational and legal perspective. The inclusion of legal feasibility is deliberate. Much of the transaction data sits inside private bilateral contracts.
The observability question
Question 1(c) is short and does more work than its length suggests. After asking what proportion of compute transactions occur at publicly disclosed prices, the Commission asks whether it would be appropriate to permit trading in a derivative contract settling to a price computed from data that the Commission may not be able to observe, verify, or surveil, in whole or in part.
That is a threshold question, not a detail. If the answer is no, a compute contract cannot settle to any index built predominantly on private bilateral data, and the product either waits for a transparent cash market or is redesigned around observable inputs.
The Commission also asks a question that assumes the answer may already exist somewhere: whether any audits, studies or independent verifications of that transaction data have been conducted, and what they concluded. Firms that have commissioned such work and never published it now have a reason to consider whether to put it into the record.
A further question asks what effect a listed contract would itself have on the cash market, including whether it would change provider incentives around publishing posted rates, structuring bilateral reservations, or disclosing utilisation and committed capacity data. The Commission is aware that listing a contract can change the market it references.
Customer protection, and a phrase worth noting
The third block covers anti-money-laundering and know-your-customer concerns, the obligations of introducing brokers and futures commission merchants, and disclosure duties.
One phrase in it is doing more than housekeeping. The Commission asks what customer-protection considerations arise from offering a compute derivatives contract settling against a geopolitically sensitive commodity, including to retail participants, and how those differ from derivatives settling against other commodities such as oil.
Calling compute geopolitically sensitive, in a document about listing standards, aligns the product with export-controlled goods rather than with ordinary industrial inputs. It also anticipates retail access, and asks separately what unique protections and prophylactic measures are appropriate for retail users of compute derivatives.
The final block asks whether perpetual compute futures would offer commercial risk management features that fixed-date contracts cannot, and what unique risks they would pose. Perpetual futures are familiar from crypto venues and unusual in CFTC-regulated markets, and the Commission is asking the question in the same breath as the compute question.
Why the CFTC frames this as an AI policy matter
The Commission grounds the exercise in its statutory duty to promote responsible innovation under the Commodity Exchange Act, and then in executive policy. It cites the July 2025 AI Action Plan and quotes its premise that whoever has the largest AI ecosystem will set global AI standards and reap broad economic and military benefits. It notes that one of the Action Plan's recommended policy actions is to ensure access to large-scale compute for startups and academics by improving the financial market for compute. It also cites Executive Order 14179, Removing Barriers to American Leadership in Artificial Intelligence.
The economic framing is drawn from a July 2026 paper the Commission cites in a footnote, which calculates that the 2025 fourth-quarter installed compute stock already implies a gross compute service flow of around 430 billion to 1.3 trillion dollars a year, or approximately 1.4 to 4.0 percent of US GDP. The Commission uses that range to support its statement that compute is a scarce, capital-intensive commodity.
The argument the Commission builds from this is that a compute futures market would let participants hedge or take price risk, and would allow financial markets to aggregate and reveal information about the future of the AI economy. That is the affirmative case. The rest of the document is a list of reasons it may not be buildable yet.
If you are contemplating a compute product, treat this as the Commission telling you what your listing filing will have to answer. The three findings to design against are in the Commission's preliminary view: compute may not yet be fungible, standardised or liquid enough. The two questions that will decide a filing are whether a contract may settle to data the Commission cannot observe, verify or surveil, and what stops a capacity provider from moving a settlement index by adjusting its own posted rate. If you administer or contribute to a compute price index, your governance, your contributor concentration and any independent verification you have commissioned are the record you will be judged on. Comments close 20 October 2026.
Source File
https://www.govinfo.gov/content/pkg/FR-2026-08-21/html/2026-17163.htm
Open the GPO text and confirm four things: the agency line and 17 CFR Parts 1 and 38 with RIN 3038-AF77 at the top, the ACTION line reading Request for comment, the DATES line setting 20 October 2026, and in Section I.c.2 the sentence stating that compute may not yet exhibit certain of the characteristics of commodities that typically underlie a commodity derivatives market.
compute may not yet exhibit certain of the characteristics of commodities that typically underlie a commodity derivatives market, including fungibility, standardization, and sufficient liquidity. CFTC, Request for Comment on the Listing of Compute Derivatives Contracts, 91 FR 54259, 21 August 2026
FAQ
Does this request for comment impose any obligation?
No. It is a request for comment, not a rule or a proposed rule. It creates no duty, amends no regulation and binds no one. Its practical weight is that it records the Commission's preliminary understanding of compute markets, and a designated contract market seeking to list a compute contract will have to address those stated views.
When are comments due?
Comments must be received on or before 20 October 2026. The document was published in the Federal Register on 21 August 2026 at 91 FR 54259.
What is the underlying commodity?
Access to computing power. The Commission describes compute as the processing power primarily used by the large language models at the centre of the AI economy, and preliminarily understands the typical underlier would be access to rented capacity from hardware the purchaser does not own, using the hourly rental price of a B200 as its example. It also raises access to a stated volume of LLM inference tokens as a possible alternative underlier.
Why does the Commission think a compute contract might not be listable?
Because of three characteristics it says compute may not yet exhibit: fungibility, standardisation and sufficient liquidity. It also states that compute markets are fragmented, that price formation occurs primarily in opaque bilateral transactions, and that dominant participants may hold pricing power that could lead to manipulability and preferential pricing.
What is the manipulation concern in concrete terms?
That published compute price series are built partly from posted rates the capacity providers themselves administer, while remaining transactions execute on venues a small number of participants dominate. The Commission asks what would prevent such a provider from moving a cash settlement index by adjusting a posted rate, steering capacity toward or away from a venue that feeds the index, or transacting selectively during the observation window.
Does this cover swaps and perpetual futures as well as exchange-listed futures?
Yes. A footnote extends the request to swap execution facilities listing contracts that settle to the price of compute and to physically settled compute swaps, and the final block of questions asks specifically about perpetual compute futures, including whether they offer risk-management features fixed-date contracts cannot and what unique risks they carry.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.