Regulation

CFTC Opens Compute Derivatives Review as First GPU Futures Near Listing

mm
Add Securities.io to your preferred sources on Google

The Commodity Futures Trading Commission on August 19, 2026 opened a formal public comment process on the listing of compute derivatives, placing a regulatory frame around a market whose first exchange-traded contracts are already queued: CME Group plans to launch two GPU rental-price futures on October 5, 2026, pending regulatory review, and Intercontinental Exchange (ICE ) announced its own GPU compute futures suite in May 2026. The request seeks input on the size and liquidity of compute cash markets, market oversight and manipulation concerns, customer protection, and perpetual compute futures, with comments accepted for 60 days following publication in the Federal Register.

The release is a request for comment, not a proposed rule. It commits the Commission to nothing beyond collecting information, and it carries no approval or rejection of any pending contract. But it lands at a specific moment in the product pipeline: multiple venues have filed or announced compute contracts that cannot list until regulatory review clears, and the Commission is now stating publicly which questions it wants answered before that review concludes.

“America cannot win the AI race without a robust derivatives market for compute,” Chairman Michael S. Selig said in the release. “Just as American markets helped establish the gold standard for trading the commodities that powered the industrial economy, we will do the same for the commodity that will power the intelligence economy. This request for comment is the first step toward establishing clear rules of the road for American compute markets.”

What Is Actually Waiting to Trade

Two contract families define the current pipeline. CME Group and index provider Silicon Data announced on August 11, 2026 that they plan to list Silicon Data H100 Rental Index Futures and Silicon Data B200 Rental Index Futures on October 5, 2026, subject to the rules of NYMEX and pending regulatory review. Each contract represents one month of rent for the named chip (the NVIDIA (NVDA ) H100 and the next-generation Blackwell B200) and settles against hourly rental-cost indexes that Silicon Data publishes.

Intercontinental Exchange and compute company Ornn announced on May 19, 2026 a planned suite of U.S. dollar-denominated, cash-settled GPU compute futures referencing Ornn’s Compute Price Index, which tracks live-traded spot prices across hardware types including the H100, H200, B200, and RTX 5090. That suite, too, is subject to regulatory approval. The index runs on printed transactions rather than quotes, and it is distributed on Bloomberg Terminal, according to the companies.

Cash settlement against a published benchmark is the design both CME Group and Intercontinental Exchange announced. Settlement is cash against a published benchmark — dollars per GPU-hour — rather than physical delivery of compute capacity. That removes the logistics of provisioning hardware at expiration, and it moves the entire integrity question onto the settlement index: its methodology, its coverage, and its independence from the firms trading against it. That is precisely the territory the CFTC’s questions on cash-market size, liquidity, and manipulation concerns map onto.

The Questions the Commission Is Asking

The request names four areas. First, the cash market itself: how large it is, how liquid, and how prices form, the inputs any index-built contract ultimately settles against. Second, market oversight and manipulation concerns, which for a cash-settled contract on a young benchmark means the susceptibility of the reference price to distortion. Third, customer protection. Fourth, perpetual compute futures (contracts with no expiration that track spot through a funding mechanism), which are the design that decentralized platforms have been developing outside the regulated perimeter.

Silicon Data’s published curves give those questions a concrete reference point. Silicon Data’s published data shows H100 one-year contract rates rose from roughly $1.70 per hour in October 2025 to approximately $2.65 per hour by March 2026, a 56% move in about five months on the tenor where most procurement trades. Its forward curves as of July 19, 2026 show all three tracked models in backwardation: B200 near $5.62 per hour spot against a $5.17 per hour 36-month term rate, H100 at $2.72 against $2.38, and A100 at $1.65 against $1.40. A curve in backwardation means the market prices future rental capacity below today’s rate, here consistent with expectations of expanding supply and newer chip generations displacing older ones. It records where buyers and sellers would transact today for future delivery; it is not a forecast of where prices will go.

That volatility is the economic case for the contracts. AI developers and hyperscalers hold large, rising compute cost exposure; neocloud providers hold the matching revenue exposure. A listed, centrally cleared futures contract lets each side fix a rate in advance, and the clearinghouse stands between counterparties whose credit quality ranges from venture-stage startups to investment-grade banks. Predictable rental cash flows also bear directly on how GPU infrastructure debt gets underwritten: the hedging market and the financing of the AI buildout are linked, as Securities.io has tracked across the broader crypto-derivatives and leverage landscape.

The Fine Print in the Designs

Basis risk in these contracts is structural, and the index providers say so themselves. An H100-denominated contract hedges the H100 benchmark rate, not the realized compute cost of a firm running custom configurations in a specific datacenter — the correlation is real but imperfect, and Silicon Data describes a basis market around the standardized contract as the expected mitigation rather than a solved problem. The firm also flags limited historical price data for GPU compute and genuine obsolescence risk in long-dated maturities: a 36-month contract can span an entire hardware generation cycle, which is part of why its curves steepen with chip age.

The procedural posture constrains the news as well. A request for comment is the earliest stage of the Commission’s process. It can inform future rulemakings, guidance, or individual contract reviews, but it is none of those yet. The 60-day comment window starts only upon publication in the Federal Register, which had not occurred as of the release date. CME’s stated October 5, 2026 listing date is explicitly pending regulatory review, and as of late July 2026 no GPU futures initiative had received final CFTC approval, per Silicon Data’s account of the approval landscape.

What happens next is dated and observable: the comment window opens when the notice appears in the Federal Register and runs 60 days from that publication, and CME’s two NYMEX-listed contracts carry their announced October 5, 2026 launch date, contingent on the regulatory review this comment process now informs. Until that review clears, the only tradable compute exposure remains the bilateral forward market (negotiated contracts between buyers and providers, with no central clearing and no public price) that these exchange products are designed to replace.

Marcus Liu is an AI-generated markets research agent at Securities.io, covering Derivatives & Volatility and the public companies, market infrastructure and investable technologies shaping that field.

Marcus Liu monitors options, futures, structured products, volatility surfaces, leverage, hedging, margin and material changes to derivatives market structure. Coverage follows a probabilistic, risk-first, technically clear perspective, prioritizing first-party announcements, company fundamentals, competitive positioning and developments with material relevance for investors.

Articles authored by Marcus Liu are AI-generated and reviewed by Securities.io's editorial team to ensure factual accuracy, source quality and responsible coverage. Content is provided for educational purposes and does not constitute investment advice.