The Pulse
CME Revises Compute Futures Effective Date to ‘To Be Announced’
CME Group has changed the effective date for its planned H100 and B200 compute futures listings to “to be announced,” after first targeting October 5, 2026. The proposed contracts would track Silicon Data GPU rental indexes, while a CFTC re
AI.info Team · Written with AI ·
CME Group has revised the effective date for its planned H100 and B200 rental-price futures to “to be announced.” The exchange and Silicon Data had announced an October 5, 2026, target date, pending regulatory review. The contracts would provide financial exposure to GPU rental benchmarks without transferring computing hardware.
CME’s October target gives way to “to be announced”
CME announced the proposed launch on August 11, setting October 5, 2026, as its target date while making the products subject to regulatory review. A revised CME clearing notice for the initial listings of the Silicon Data H100 and B200 Rental Index Futures now gives the effective date as “to be announced.” CME’s product page describes the launch as planned for later this year and pending regulatory review. The documents do not state why the effective date is unannounced.
The Commodity Futures Trading Commission issued a public request for comment on compute derivatives on August 19. The agency sought feedback on the size and liquidity of compute cash markets, market oversight and manipulation concerns, customer protection, and perpetual compute futures. It said comments would be accepted for 60 days following publication in the Federal Register. The request is part of the regulatory context for the planned products, but the CFTC announcement does not say that it caused CME to change the effective date.
Two GPU benchmarks, one-month contract units
The planned contracts track Silicon Data indexes measuring hourly on-demand rental costs for Nvidia H100 and Blackwell B200 GPUs. CME identifies the products as GPU1 and GPU2. Each contract represents 730 GPU-hours, which CME describes as one month of capacity, and is financially settled. Monthly contracts are listed out to 36 months.
The indexes cover neocloud providers, excluding hyperscalers, according to CME. That scope gives the contracts a defined reference price, but it may not match the rates or terms a particular buyer negotiates. The difference between a company’s actual rental bill and the index could limit how closely a contract offsets its costs. Because the contracts are financially settled, they provide exposure to price movements rather than GPU access.
Silicon Data says the benchmark is the point
When the companies announced the plan, Silicon Data Chief Executive Officer Carmen Li said:
“Compute futures give the market something it's never had: a public, tradable reference price for the resource every AI system runs on.”
CME said the contracts could help AI developers, hyperscalers and other businesses manage price risk in compute spending. The August announcement described them as a potential way for companies to lock in costs and gain a window into future AI spending. The contracts would provide exposure to index movements, not reserve GPU capacity or guarantee rental prices from any provider.
The October 5 date was a proposed start date, not confirmation that trading had begun. CME’s revised notice leaves the effective date open, and the exchange’s product page continues to mark the launch as pending regulatory review. The H100 and B200 contracts therefore remain planned listings, rather than available hedges.