NVIDIA launches revenue-sharing model for AI cloud GPU deployments, shifting from pure hardware sale to cloud-services partnership.

On July 1, Nvidia launched a revenue-sharing and credit-support model for AI cloud operators, enabling them to deploy Grace Blackwell GB300 GPUs without requiring full upfront capital. Under this structure, Nvidia earns standard hardware revenue plus a recurring share of the cloud income generated by those GPUs.

The move represents a fundamental shift in how the world's most valuable semiconductor company generates revenue. GPU sales alone are no longer sufficient. Nvidia now takes a share of the revenue that those GPUs produce, betting that AI cloud is the future and positioning itself to capture value across the entire ecosystem.

The model addresses a critical pain point for cloud operators. If they deploy Nvidia GPUs but cannot fully utilize compute slots, Nvidia backstops the gap by renting or purchasing idle capacity at predetermined prices. This means Nvidia shoulders risk alongside its partners, rather than simply capturing upfront margins.

The first partners under this arrangement are Sharon AI and Firmus Technologies, which have jointly committed to 210,000 Grace Blackwell GB300 GPUs. Sharon AI is deploying 40,000 GPUs across a 72-megawatt facility in Australia. Firmus is constructing a 360-megawatt facility in Indonesia, with plans to scale to 170,000 Nvidia GPUs.

Cloud operators have historically struggled to finance multi-billion-dollar buildouts, needing GPUs immediately but lacking capital until revenue streams materialize. By introducing this model, Nvidia has positioned itself as an infrastructure partner, not merely a vendor, solving a structural financing problem.

This approach could democratize AI cloud deployment. Smaller operators unable to afford $10 billion in upfront GPU purchases can now deploy at scale more quickly. Nvidia's calculation is straightforward: greater long-term value lies in capturing revenue share from thousands of successful deployments than in requiring cash upfront from a smaller number of well-capitalized buyers.

Revenue sharing inherently depends on operator success. If deployments falter or remain idle longer than anticipated, Nvidia absorbs the losses. The company is effectively financing its own customers—a profitable arrangement if they thrive, but costly if they fail. The trade-off is clear: faster ecosystem growth but lower margins per dollar deployed. Over the next two years, this bet will prove either prescient or reckless.