Nvidia Moves Beyond GPUs
Nvidia is no longer just selling GPUs—it is underwriting the next wave of AI cloud infrastructure.
The company has introduced a revenue-sharing model designed to lower the barrier to entry for AI cloud startups. Instead of absorbing the full upfront cost of GPU infrastructure, partners can deploy Nvidia hardware with financing tied to future compute revenue.
A New Risk Model for AI Compute
What stands out is how Nvidia is reshaping risk.
Partners share revenue generated from AI workloads, while Nvidia absorbs part of the downside. If deployed capacity goes underutilized, Nvidia agrees to buy back unused GPU inventory under pre-defined terms. This effectively positions Nvidia as more than a hardware vendor—it becomes a hybrid of infrastructure partner, financier, and market maker for AI compute.
First Deployments at Massive Scale
The initial partners signal the scale Nvidia is targeting.
Sharon AI has committed up to $4.88 billion over six years to deploy tens of thousands of Grace Blackwell GPUs across 72 megawatts of capacity in Australia. Firmus Technologies is going further, building a 360-megawatt AI campus in Indonesia, with plans to scale to 170,000 accelerators across multiple Nvidia architectures.
What This Means for Founders
For founders and infrastructure builders, this model changes the equation.
Access to high-end compute has been one of the biggest bottlenecks in building AI-native platforms. Nvidia is now offering a path to launch AI cloud infrastructure without hyperscaler-level capital—provided there is real demand behind it.
The tradeoff is dependency. Startups entering this model are tying their economics and growth to Nvidia’s ecosystem, pricing, and roadmap.
Nvidia’s Expanding Control of the Stack
Nvidia is also extending its reach beyond infrastructure.
The company has taken equity stakes in some partners and is onboarding inference platforms such as Baseten, Fireworks AI, and Together AI. This signals a deeper push into owning not just the hardware layer, but also the platforms and services running on top.
The Bigger Shift
The first cohort represents more than 430 megawatts of AI capacity expected to come online starting in 2027.
If this model proves successful, Nvidia will not just supply the infrastructure powering AI—it will help finance it, shape it, and increasingly control how it is consumed.
