Kalshi’s CEO is racing to build a futures market for AI’s most precious resource, which could be worth $100 trillion by 2030

When the price of jet fuel skyrocketed at the onset of the Iran war, airlines like Lufthansa avoided financial turbulence by purchasing hedging contracts that ensured most of their fuel purchases would remain locked in at pre-war prices. Today, businesses that heavily rely on compute resources, often referred to as the new oil, are eager for a similar hedging option.

They may soon have one. According to Kalshi CEO Tarek Mansour, compute—the term for the chips and electricity driving the AI revolution—will surpass oil as the world's most valuable commodity. If compute follows the pattern of derivatives markets for other commodities, its futures market could grow to 10-15 times the size of the underlying spot market, potentially reaching $100-$150 trillion.

Kalshi CEO Tarek Mansour's prediction presents a massive opportunity for whoever can establish this market. In July, Kalshi unveiled a series of new event contracts and data tools that could serve as the foundation for a compute derivatives market. However, building such a market is no easy feat, and the history of commodities markets reveals the process can be slow, complicated, and uncertain.

To hedge compute costs, companies face a challenge similar to what oil market participants encountered in the 1970s. Oil market participants demanded a futures market to hedge against sudden price fluctuations, but financial institutions struggled to agree on a standard barrel size. Similarly, AI compute costs fluctuate significantly, with hourly prices sometimes varying by up to 137% annually.

This volatility is exacerbated by the limited standardization of AI compute pricing, which relies on B2B contracts between suppliers like Nvidia and corporations like HP. Additionally, newer chips lead to higher efficiency, but the degree of improvement is difficult to predict, further complicating the effort to forecast future compute prices.

Despite these challenges, Kalshi believes prediction markets can help establish compute futures. Currently, Kalshi offers wagers on five types of chips, with bettors predicting whether the average hourly cost of renting Nvidia's H200 chip in August will be above or below $5. Kalshi relies on Ornn to publish a popular dashboard showing the cost of renting various hardware, which Kalshi uses to create forward curves predicting price movements up to a year in advance.

While Kalshi's compute futures market currently pales in comparison to the oil futures market, the potential for growth is undeniable.