# CME Group Sets October 5 for Futures on Nvidia GPU Rental Prices

Two cash-settled contracts tied to Silicon Data's H100 and B200 hourly rental indexes will list on the New York Mercantile Exchange, each representing one month of rent for a single accelerator.

- Published: 2026-08-23T06:19:43.709Z
- Canonical: https://polylog.news/ai/2026-08-23/cme-group-sets-october-5-for-futures-on-nvidia-gpu-rental-pr
- Publisher: Polylog (AI desk)
- Section: markets
- Sources: [Polylog editors](https://polylog.news), [CME Group](https://www.cmegroup.com/media-room/press-releases/2026/8/11/cme_group_and_silicondatatolaunchcomputefuturesonoctober5tounloc.html), [The Next Web](https://thenextweb.com/news/cme-silicon-data-gpu-compute-futures-october-5)

CME Group and the data provider Silicon Data will begin trading [two compute futures contracts on October 5](https://www.cmegroup.com/media-room/press-releases/2026/8/11/cme_group_and_silicondatatolaunchcomputefuturesonoctober5tounloc.html), subject to regulatory review. One settles against Silicon Data's index of hourly rental rates for the Nvidia H100, the other against the equivalent index for the Blackwell B200. Both list on the New York Mercantile Exchange (NYMEX) and clear through CME Clearing, and each contract corresponds to a month of rent calculated from the hourly rate.

The buyers CME has in mind are the parties on both sides of the rental market. Model developers and enterprises that lease capacity can fix a forward cost. Data-center operators and neoclouds that own the hardware can lock in revenue against a fixed financing schedule. A [Russian-language technical channel summarizing the launch](https://t.me/ai_machinelearning_big_data/10746) framed the appeal the same way, as a hedge against rental-price swings that currently have no instrument behind them.

The important effect is informational, not financial. Today, GPU rental pricing is quoted privately between the parties to each deal and disclosed selectively, so nobody outside a given contract knows what capacity actually costs. A cleared futures curve publishes a forward price instead, which means traders will set a visible price reflecting whether they expect Blackwell capacity to be scarce or in surplus six months out. [Coverage of the launch](https://thenextweb.com/news/cme-silicon-data-gpu-compute-futures-october-5) has focused on that transparency point.

Two caveats apply. Settlement depends on Silicon Data's index methodology, and an index built from a thin or unrepresentative sample of rental quotes can be gamed or can simply track the wrong thing. And a new contract only becomes a useful hedge if it attracts liquidity, which most commodity launches fail to do in their first year.

## What this means

A cleared forward curve turns compute from a procurement line item into a priced commodity, which changes who bears the risk of an accelerator glut. Neoclouds and GPU leasing firms gain the ability to pre-sell capacity and finance hardware against contracted revenue, which lowers their cost of capital. Nvidia's customers gain a public benchmark that makes it harder for suppliers to price-discriminate quietly. The exposed party is any operator whose business model assumes rental rates stay elevated, because a visible forward curve, whether it prices future months above or below today's rate, will now make that assumption visible to the rest of the market.

## What to watch

- Whether open interest and daily volume in the two contracts build past the launch weeks, because a contract nobody trades cannot function as a hedge for anyone.
- How the published forward curve for B200 rentals compares with H100, since a steep discount on the older chip would give the first market-based reading on how fast accelerator generations lose value.
- Whether Silicon Data publishes its index construction and contributor set in enough detail for participants to audit it, which decides how much trust the settlement price can carry.