NVIDIA is moving to a new model - sharing revenue with AI clouds, not just selling them hardware. Sharon AI (an AI cloud services company) is heading toward 40,000 GB300 GPUs; Firmus (an AI data center company) is building a 360-megawatt campus in Indonesia.
- NVIDIA announced a revenue-sharing model (July 1): it takes a cut of cloud services, not just hardware revenue.
- Sharon AI is deploying up to 40,000 Grace Blackwell GB300 GPUs; Firmus is building a campus up to 360 MW and 170,000 GPUs in Indonesia.
- The logic: new AI clouds don't have capital for chips at this scale, and NVIDIA has an interest in them existing.
NVIDIA announced a revenue-sharing model with its cloud partners. Besides revenue from hardware, the company now also takes a cut of the GPU-based cloud services built on the capacity it itself helps build. The logic is simple: startups and new AI clouds don't have the capital for GPUs at this scale, and NVIDIA has an interest in them existing.
This is bigger news than another chip. NVIDIA is moving from 'shovel seller' to 'co-owner of the mine.' Once a supplier takes a percentage of the service, it wants that service to run long-term - not just to sell you hardware and walk away.
AI infrastructure got expensive enough that even its buyers can no longer buy it alone. What's left is sharing it. Watch who owns what - that's where power will shift over the coming years.