AI Pulse · Business Models & Winners

CME Group and Silicon Data plan to launch cash-settled futures contracts tied to GPU rental prices for Nvidia’s H100 and B200 chips on October 5, subject to regulatory review. Put simply, companies will be able to trade contracts based on where GPU rental prices are expected to go, rather than only buying compute at whatever price is available on the day.

That matters because AI compute is becoming a market with a clearer price signal. Today, many operators struggle to budget for serious AI workloads because capacity and rental rates can move quickly. A recognised index, and futures contracts built around it, gives buyers, providers and investors a common reference point. It also creates a way to manage exposure to future compute costs, much like businesses use contracts to plan around energy, currency or commodity prices.

There’s already evidence that these rates are being treated as meaningful market data. SpaceX’s own SEC filing cited Silicon Data’s median H100 rental rate of $3.33 per hour when sizing a $2.4 trillion compute infrastructure opportunity. That doesn’t make the estimate certain, but it does show that GPU pricing is moving from an opaque vendor conversation into financial planning and investment models.

For business owners, the practical takeaway is simple. Compute cost is becoming a line item you’ll need to watch, forecast and potentially hedge as AI use grows. This is the kind of thing we build into an AI command centre, connecting AI activity to the operating numbers that affect real decisions.