Nvidia announced on July 1 a new revenue-sharing and credit-support financing model that lets smaller AI cloud providers deploy massive GPU fleets without paying the full cost upfront. Instead of one-time hardware sales, Nvidia backstops the financing needed for data-center buildouts and takes a declining share of the cloud revenue those chips generate over the life of the contract.
The first two partners are Australia's Sharon AI, deploying up to 40,000 Grace Blackwell GB300 GPUs, and Firmus, which is building a 360-megawatt "DSX" AI factory campus in Batam, Indonesia, scaling to as many as 170,000 GPUs. Combined, the two deals unlock roughly 210,000 GPUs at hyperscaler-class scale.
The model targets inference-focused platforms such as Baseten, Fireworks AI and Together AI as the industry shifts from training frontier models to running them continuously in production. Critics warn the arrangement shifts financing risk onto Nvidia's balance sheet, but the company frames it as a way to get more of its chips into the hands of capital-constrained AI startups.