Nebius and Crusoe say AI compute rental prices are rising despite debt fears

AI computing capacity is fetching higher rental prices, executives at Nebius and Crusoe told CNBC in interviews published October 21, 2026. Nebius said prices were beating its expectations, while Crusoe pointed to stronger rents for older Nvidia chips—a counterpoint to investor concerns about the debt financing the AI infrastructure buildout.

The companies operate in the neocloud market: businesses that build data centers, buy graphics processing units, or GPUs, and rent computing capacity to customers. Providers spend billions on hardware that must keep earning revenue to recover those investments and generate profits.

Investors have been scrutinizing debt, borrowing costs and data-center economics after CoreWeave’s cost of borrowing jumped. That pressure has spread beyond CoreWeave, with Nebius and Iren shares falling sharply over the past month.

Nebius points to an auction, not just an outlook

Nebius Chief Revenue Officer Marc Boroditsky said a computing-capacity auction held October 20 showed continued increases in price per megawatt. That measure describes what providers can charge for a given amount of computing capacity. Asked whether it was still rising ahead of expectations, he answered yes.

Boroditsky also said Nebius would sell out its 2027 capacity if it brought that capacity online. He described a similar demand dynamic as possible in 2028. Those are conditional projections, not confirmation that all the planned capacity has already been delivered or sold.

On financing, Boroditsky distinguished Nebius from CoreWeave, saying his company was not carrying the same kind of debt.

Crusoe’s case rests on older chips

Crusoe CEO Chase Lochmiller addressed a different concern: how quickly expensive chips lose their earning power when newer generations arrive. He said older Nvidia hardware was commanding higher rents at Crusoe—a company-specific comparison, not a market-wide rental-price measure.

Lochmiller attributed strong demand to advances in AI models, arguing that customers can now create more value from the underlying computing infrastructure.