H100 rental prices slide after early May surge, says Ornn AI Inc

The cost of renting Nvidia’s H100 GPUs is coming back down after a spike earlier this month, according to data from the Ornn Compute Price Index. For anyone running AI workloads on rented hardware, the timing matters: the GPU rental market is no longer just an infrastructure concern. It’s becoming a full-blown financial market.

Ornn AI Inc., which publishes the OCPI, a benchmark derived from actual trades rather than listed prices, has been tracking the pullback in real time. The index is now accessible on the Bloomberg Terminal, putting GPU rental pricing alongside oil futures and Treasury yields in the toolkit of institutional traders.

The numbers behind the slide

H100 rental rates have been on a wild ride. To put the trajectory in context, hourly rates hit around $8 back in early 2024, when demand for Nvidia’s chips far outstripped supply. By March 2026, one-year contract pricing had settled closer to $2.35 per hour. Recent prediction market activity pegged rates near $1.80 per hour.

Then came the early May surge, pushing prices back up before the current decline set in. The OCPI captures these moves because it tracks live-traded prices, not broker quotes or self-reported figures.

The derivative side of the market tells a similar story. Ornn’s H100 Compute derivatives product hit an all-time high of $2.94 on May 25, 2026. Shortly after, it pulled back by roughly 3% to 9%.

Other indices have cited H100 prices ranging from approximately $1.70 to $2.63 per hour, depending on contract length and provider.

GPU compute gets its own futures market

On May 19, 2026, ICE and Ornn announced they would launch GPU compute futures contracts based on the OCPI. The contracts will cover multiple GPU models, including both the H100 and the newer H200.

For AI companies renting thousands of GPUs at a time, the ability to hedge compute costs could meaningfully change how they budget and plan. For financial traders, it creates an entirely new asset class to speculate on.

Tens of billions of dollars are currently flowing into AI infrastructure globally. The OCPI and its associated derivatives are essentially an attempt to build financial rails around that capital.

What this means for investors

The GPU rental market is heavily dependent on Nvidia hardware, which means a single company’s production schedule, export restrictions, and product roadmap can move the entire market. Early volume and open interest figures from the ICE futures contracts will reveal whether this market has genuine two-sided participation or is still dominated by a handful of large players. The spread between spot OCPI rates and futures prices will also be telling: a steep contango would suggest the market expects prices to rise, while backwardation would signal the opposite.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.