Kalshi bets compute prices can be forecast a year ahead

AI labs are scrambling for compute, and Kalshi reportedly wants to sell them something adjacent: a way to forecast where compute prices are heading. According to Bloomberg, the prediction market company is developing a tool that maps the expected future price of computing power, potentially giving customers a chance to lock in capacity before costs rise further.

That idea lands at a moment when compute remains scarce. Demand is still outpacing the data center buildout needed to support it. Former Intel CEO Pat Gelsinger recently told CNBC that demand is “almost unlimited.” The constraint, of course, is that available energy and processing capacity are not.

A recent report from Apollo Global Management described current compute capacity as “effectively sold out,” creating a bottleneck in which GPU rental prices are climbing faster than new supply can come online. The pressure could intensify as more agentic AI systems are deployed; recent research suggests they can consume up to 136.5 times more energy per query than most generative AI models.

Per Bloomberg, Kalshi’s tool would provide a forward tracking curve for compute prices. It would analyze weekly and monthly contracts for computing power and use an algorithm to estimate where prices are likely to land, with forecasts reportedly stretching as far as one year out.

Bloomberg did not say whether Kalshi plans to turn that forecast into a tradeable market where users can bet on whether actual prices come in above or below the predicted level. But other exchanges are already looking at compute futures contracts, treating access to processing power as a financial asset. If that market takes shape, compute availability may soon be something companies hedge—and speculators trade.

Recommended reading

AI Coding Agents Keep Babel Standing