Nvidia pauses AI cloud revenue-share plan

Nvidia has paused new deals under its AI Compute Partnership, a financing program launched less than two months ago to help smaller AI cloud providers afford GPU capacity, according to the Wall Street Journal. Under the program, Nvidia agreed to rent back computing capacity from a cloud provider if that provider could not sell it themselves, in exchange for 50 percent of any revenue above an agreed hourly baseline covering the provider’s costs.

Two problems reportedly drove the pause. Some Nvidia employees warned internally that the arrangement, which required cloud providers to route customers through an Nvidia approval process and spread utilization across several smaller firms rather than concentrating it with one large buyer, could draw antitrust scrutiny. Separately, the control terms irritated prospective partners during the program’s early weeks. Nvidia has denied outright pausing the initiative, per a later Tom’s Hardware report, but the Journal’s sourcing says the company stepped back from active deals last week and could revamp or fold the program into something else later.

The bind smaller clouds are back in

The program existed specifically to solve a financing gap: smaller AI cloud operators often cannot get bank debt for GPU purchases without a buyer commitment, which is exactly what Nvidia’s revenue-share was meant to backstop. Its pause lands the same week Lambda secured $1 billion in separate debt financing to buy Nvidia chips outright rather than lean on vendor financing, one path around the gap while Nvidia’s own program is on ice. For any smaller cloud provider counting on this kind of backstop, the near-term message is that Nvidia’s appetite for taking on partner-concentration and antitrust risk directly has cooled, even as GPU demand keeps climbing.