Overview
- Nvidia announced the new AI Compute Partnership publicly on July 2 and named Sharon AI and Firmus as the program’s first partners with multi‑year, multi‑megawatt DSX deployments.
- The program lets AI cloud providers access Nvidia hardware and token credits now while Nvidia collects its usual product revenue plus a recurring percentage of the cloud income those GPUs generate.
- Sharon AI disclosed in an SEC filing a six‑year, 72 MW agreement that can scale to about 40,000 Nvidia Grace Blackwell GB300 GPUs and Firmus plans a DSX campus in Batam targeting 360 MW and roughly 170,000 GPUs.
- Key commercial terms such as the revenue‑split percentages remain undisclosed and analysts warn the model creates circular financing, stacked obligations for smaller clouds and exposure if utilisation falls short.
- The initiative turns Nvidia’s DSX design into a repeatable financing product that could speed regional AI capacity buildout, and its proof points will be partner filings, utilisation rates and commercial disclosures over coming quarters.