NVIDIA is testing a different way to finance large AI cloud deployments: sell the hardware, but also share in the cloud revenue tied to the supported capacity.

In a July 1 blog post, the company said it is working with AI cloud providers on a revenue-sharing and credit-support model for NVIDIA-powered "AI factories." The pitch is aimed at startups, model builders, inference providers, enterprises, and research users that need accelerated compute without waiting for a full data center buildout.

The structure is notable because it moves NVIDIA a step beyond normal chip sales. Under the model described by the company, AI clouds sell NVIDIA-powered services, while NVIDIA receives standard product revenue plus a share of the cloud revenue on the supported capacity.

Two early deployments show the scale NVIDIA is targeting. Sharon AI said its six-year collaboration is designed to enable 72 megawatts of new data center capacity in Australia and up to 40,000 Grace Blackwell GB300 GPUs. Firmus said it is building a NVIDIA DSX AI factory campus in Batam, Indonesia, with planned capacity of 360 megawatts and up to 170,000 GPUs.

The announcement should still be read as an infrastructure financing model, not proof that all of that capacity is online. The verified change is that NVIDIA is using its balance sheet and platform position to help more AI cloud operators finance Blackwell-era capacity, while tying part of its upside to actual cloud usage.