NVIDIA Disrupts Tech Supply Chain With New GPU Revenue Sharing Model

NVIDIA is trying something new in the AI infrastructure race, and it goes beyond simply shipping more hardware. The company has rolled out a credit support and revenue sharing model, that could reshape how emerging AI cloud providers get their hands on serious amounts of compute. In plain terms, NVIDIA still gets paid upfront when a cloud provider buys the hardware, but it also collects an ongoing share of the cloud revenue that hardware generates once it's up and running, in a kind of two-for-one on the same stack of GPUs.

This is a significant departure from the old playbook. Traditionally, a company bought the processors, built the data center, rented out the compute, and hoped the math worked out eventually. NVIDIA's new arrangement ties the company far more closely to how that infrastructure performs after the sale closes. To paraphrase NVIDIA's own description, where the old model waited for someone else to strike gold, this one lets the company hold a permanent stake in the mine.

The timing tracks. Companies now want always-on AI factories capable of grinding through training, fine-tuning, post-training, and enormous volumes of inference. That last piece is particularly important in today's market, since inference is a massive growth area at the moment. More users, more autonomous agents, and more AI-powered apps translate directly into demand for more compute.


That kind of demand creates a serious financing problem for some companies. Standing up AI infrastructure eats enormous amounts of upfront capital, and smaller AI outfits or specialized cloud providers cannot always unlock funding on the strength of a customer's promise alone. NVIDIA says this model is meant to close that gap, backing AI cloud partners with credit support while sharing in the revenue those partners pull in from NVIDIA-powered cloud services.

The opening lineup is already substantial. Sharon AI plans to deploy as many as 40,000 Grace Blackwell GB300s. Sharon AI co-founder and CEO James Manning called the NVIDIA partnership a pivotal step toward building sovereign, large-scale AI compute. Firmus Technologies, meanwhile, is constructing a DSX AI factory campus in Batam, Indonesia, expected to scale to 360 megawatts and as many as 170,000 NVIDIA GPUs, with co-CEO Tim Rosenfield framing scalable, efficient compute as essential for AI-native companies competing globally.

NVIDIA also points to Baseten, Fireworks AI, and Together AI as the kind of AI-native companies driving this new model. For NVIDIA, the strategic upside is obvious. This model widens the market for its GPUs while building a recurring revenue stream tied to actual cloud usage, embedding the company deeper into its partners' business models. It also means the tech titan is no longer only a component and solution supplier. It's becoming a financial and infrastructure partner for AI cloud providers, one GPU-packed factory at a time.