The CFTC has sent a request for public comment on CME compute futures to the White House Office of Management and Budget for review, a step that could push back the October 5 launch date for the two GPU-linked contracts CME Group and Silicon Data announced in August.
Bloomberg reported the regulatory move on Monday. Once the White House review concludes, the CFTC is expected to open a public comment period of either 30 or 60 days, meaning the clock on any approval is now running on a timeline set by Washington, not Chicago.
What the CME Compute Futures Actually Look Like
The two contracts are the Silicon Data H100 Rental Index Futures (Globex code: GPU1) and the Silicon Data B200 Rental Index Futures (Globex code: GPU2), both to be listed on NYMEX pending approval. Each contract unit is 730 GPU-hours, representing one month of rental capacity. Both are cash-settled, with a tick value of $7.30, and monthly contracts are listed out 36 months.
The benchmark pricing comes from Silicon Data, which publishes daily on-demand rental rates for H100, A100, and B200 chips, along with a GPU forward curve the firm describes as the first of its kind. Silicon Data is backed by global trading firm DRW; its CEO is Carmen Li.
At the August 11 launch announcement, Li made the pricing-opacity problem explicit: ‘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. They will now have a benchmark to check that against.’
The original partnership between CME Group and Silicon Data was announced on 12 May 2026, with CME Group Chairman and CEO Terry Duffy framing compute as ‘the new oil of the 21st century’ and adding: ‘Every AI model trained, every transaction cleared, and every byte of data processed runs on compute, which is becoming a fast-emerging asset class in its own right.’
Forward Curve Signals Where the GPU Market Is Heading
Silicon Data’s published H100 forward curve as of 7 September 2026 shows the market pricing in a softening of rental rates. The three-month forward sat at approximately $2.50, declining to $2.14 at twelve months out, with the 12-month strip averaging $2.26 against a spot rate of $2.63.
That contango-in-reverse structure matters for anyone planning to use the contracts as a hedge rather than a directional trade. A hyperscaler or AI lab locking in GPU capacity costs twelve months forward would be doing so below current spot, which is a useful data point for capital allocation decisions even before the first contract clears.
Note that contract terms remain subject to pending CFTC review; the figures above are from Silicon Data’s practitioner’s guide and reflect market conditions as of that date.
CME Is Not the Only Exchange Eyeing This Market
The regulatory queue is filling up. Intercontinental Exchange and Ornn announced plans on 19 May 2026 to launch US dollar-denominated, cash-settled GPU compute futures based on Ornn’s Compute Price Index. A third entrant, Architect, announced its own compute futures on a new US-based exchange on 28 May 2026, according to a compute derivatives market primer published on Substack.
All three products sit in the same regulatory waiting room. The CFTC’s decision to solicit public comment before approving any of them suggests the agency wants to establish a framework rather than wave through contracts one by one.
The backdrop is a GPU market that has already been moving. CNBC reported that memory prices soared in the first quarter of 2026, helping chip manufacturers diverge from the broader equity market. Estimates from TD Lombard, Goldman Sachs, and Bridgewater Associates, cited in the original reporting, put AI infrastructure spending at roughly 2% to 2.5% of US GDP this year.
The CFTC comment period, once opened, is where the real negotiation happens. Market structure questions around settlement methodology, benchmark manipulation risk, and position limits will all surface. How the agency resolves them will determine whether October 5 is a launch date or a placeholder.
