The US Commodity Futures Trading Commission has sent a draft request for public comment on artificial intelligence compute futures to the White House Office of Management and Budget, Bloomberg reported on Monday. The review lands seven weeks before CME Group’s planned Oct. 5 launch of two contracts priced off GPU rental benchmarks. Compute futures are contracts that let a buyer lock in the future cost of renting graphics processing units, the hardware that trains and runs AI models. The timing carries weight for bitcoin miners, many of which have converted power sites into AI hosting capacity and would be selling into the market these benchmarks are built to price.
Key takeaways
- The CFTC sent a draft request for comment on compute futures to the White House OMB, according to an Aug. 17 Bloomberg report.
- CME Group plans to list Silicon Data H100 Rental Index Futures and Silicon Data B200 Rental Index Futures on NYMEX on Oct. 5, pending regulatory review.
- No request had appeared on the CFTC’s comment pages or in the Federal Register as of Aug. 18, leaving the questions and deadline unconfirmed.
- Intercontinental Exchange is developing competing compute contracts using separate benchmarks, splitting the field before any market opens.
Published: Aug. 18, 2026, 16:30 UTC
A comment request is not a rejection
A request for comment is not a proposed rule, and it does not by itself stop CME’s Oct. 5 date. Registered exchanges can bring new products to market through self certification, a process that lets an exchange list a contract after filing terms with the CFTC rather than waiting for approval. The agency retains authority to review those terms and act where the law permits.
What the request signals is unsettled ground. The CFTC has not publicly confirmed any delay. But a consultation broader than a single contract filing suggests the agency wants input on questions that apply to the whole category: whether the underlying benchmarks resist manipulation, how settlement should work, and how an exchange defines a standardized unit of computing power.
As of Aug. 18, nothing had been published. Only the released document can confirm the questions and the response window, which Bloomberg reported would likely run 30 or 60 days once OMB finishes its review.
What CME plans to list on Oct. 5
CME Group and Silicon Data announced on Aug. 11 that two contracts would list on NYMEX on Oct. 5, subject to regulatory review. Both track daily indexes of hourly GPU rental costs published by Silicon Data, a GPU market intelligence firm backed by trading firm DRW. One contract represents a month of rent for Nvidia’s H100. The other covers the next generation Blackwell B200.
The pitch is price transparency in a market that has not had it. “For years, two companies buying the exact same GPU capacity could pay wildly different prices with no way to know who got the better deal,” Silicon Data CEO Carmen Li said in the announcement. GPU rental costs vary by hardware, provider, region and contract length, which makes comparison difficult and hedging close to impossible.
CME frames the product in commodity terms. Pete Keavey, the exchange’s global head of energy and environmental products, said the contracts would “turn compute into a standardized, tradable commodity.” That is a market description, not a regulatory classification, and the distinction is the point of the CFTC’s review.
Why bitcoin miners have a stake in the benchmark
The crypto industry sits on the supply side of this trade. Miners hold what AI operators need most: grid interconnection, power contracts, cooling systems and large industrial sites. Several have moved decisively. Riot Platforms signed a $9.1 billion AI lease reported to be with Anthropic. TeraWulf reported more revenue from AI hosting than from bitcoin mining in the first quarter of 2026. Galaxy Digital delivered 133 megawatts to CoreWeave under a 15 year agreement at a former mining campus in Texas.
Those deals were negotiated privately, one counterparty at a time. A liquid futures curve would change that. It would give miners a public forward price to benchmark lease terms against, and give their investors a way to value contracted capacity that does not depend on management disclosure. It would also let operators hedge revenue on capacity they have not yet leased, which is closer to how the sector’s $70 billion AI pivot has been financed than to how bitcoin mining has traditionally been run.
Whether that materializes depends on liquidity, and liquidity is where the competing benchmarks cut against the industry. Intercontinental Exchange is building cash settled contracts on Ornn’s Compute Price Index, which tracks the H100, H200, B200 and RTX 5090, plus a separate product using NativX’s COIL Index for energy normalized compute. More benchmarks give users choice. They also divide trading activity across incompatible reference prices.
What comes next
Three things have to happen before compute trades. OMB must finish its review and the CFTC must publish the request. CME must complete its contract filing. And enough participants have to show up to make the curve meaningful.
The macro case is not in doubt. Estimates from TD Lombard, Goldman Sachs and Bridgewater Associates put AI infrastructure spending near 2% to 2.5% of US GDP this year, a level that generates real hedging demand. The open question is the one the CFTC is now asking: whether a GPU rental index is a durable enough reference to settle regulated futures against, particularly as newer chips displace the hardware the index was built on. Oct. 5 remains a target, not a confirmed trading date.
Frequently asked questions
What are AI compute futures?
They are futures contracts tied to the cost of renting GPU computing capacity. Buyers use them to lock in or hedge future compute costs, and traders can gain exposure to compute prices without owning chips or running a data center. CME’s versions settle against daily GPU rental indexes rather than physical delivery.
Does the CFTC review block CME’s October 5 launch?
Not automatically. A request for public comment is not a proposed rule or a final decision, and CME can list products through self certification. The launch would only slip if the CFTC separately objects to the contracts or requires further review. CME has said all along that the date depends on regulatory review.
How does this connect to bitcoin mining?
Bitcoin miners control power capacity and industrial sites that AI operators want, and several now earn more from AI hosting than from mining. A public compute price curve would let them benchmark lease negotiations and hedge revenue on uncontracted capacity, replacing private bilateral pricing with a market reference.
