Nvidia Launches AI Revenue-Sharing Model for Startups and Cloud Services

Nvidia unveiled a new revenue-sharing program this week that lets AI startups, model builders, and cloud providers access its computing infrastructure without financing the hardware buildout themselves. Under the model, participating cloud providers deliver NVDA-powered services and share a cut of their future cloud earnings with Nvidia on top of standard hardware sales, while startups can draw on token credits instead of committing upfront capital.

Sharon AI and Firmus are among the first partners in the program. Sharon AI plans to deploy as many as 40,000 Nvidia Grace Blackwell GB300 GPUs under the new arrangement, and Firmus is developing a campus in Batam, Indonesia designed for up to 170,000 Nvidia GPUs and 360 megawatts of power. The structure removes the need for smaller AI developers to navigate site selection, power procurement, construction, and hardware bring-up before accessing large-scale compute.

The move extends Nvidia's push beyond one-time chip sales into recurring, usage-linked revenue, building on similar compute-financing arrangements the company has struck with major cloud vendors over the past year. By tying its own earnings to the success of the startups and clouds it supplies, NVDA effectively subsidizes AI infrastructure buildouts for smaller players while retaining a claim on the upside if their businesses scale, a structure that could deepen customer lock-in but also ties Nvidia's results more closely to its partners' execution risk.