NVIDIA is changing the way AI infrastructure is financed and deployed, unveiling a new partnership model with AI cloud providers that aims to make high-performance computing more accessible as demand for artificial intelligence continues to surge.
The company said it is partnering with AI cloud operators to build large-scale, multi-tenant AI factories, supported by a revenue-sharing and credit-support model. Instead of merely selling hardware, NVIDIA will get a cut of the revenue from the AI services running on its infrastructure. A model that aligns the interest of both parties.
The announcement comes as the AI industry is entering a new phase. Most of the focus, until recently, has been on training large AI models. Now the industry is moving to AI inference, the phase where trained models generate real-time responses for users.
Why is the deal significant?
Every chatbot conversation, every AI-generated image, every coding assistant request requires computing power, and that demand is growing fast as AI applications proliferate.
To support that shift, companies need computing infrastructure that can run 24/7 and serve multiple customers at once. NVIDIA calls these facilities AI Factories — specially-designed data centers for handling massive quantities of AI workloads at scale.
But building that sort of infrastructure is expensive. And for many startups and upstart AI cloud providers, raising the capital to deploy massive GPU clusters has been a huge challenge, even if they have customers standing in queue. NVIDIA thinks its new business model can help bridge that gap.
NVIDIA will help with deployments by providing a mix of financing and revenue sharing, rather than requiring cloud providers to pay for all the financial costs up front.
AI cloud providers will provide startups, enterprises, software vendors and AI developers access to computing services powered by NVIDIA. NVIDIA will generate revenue from the hardware it provides and a portion of the cloud revenue generated from the supported capacity.
The approach is designed to make it easier for cloud providers to scale and gives NVIDIA a recurring revenue stream based on customer usage, not one-time hardware sales.
Benefits of the deal
Rather than years spent on land acquisition, power infrastructure, construction and hardware installation, developers would have faster access to ready-to-use AI computing resources. It enables them to concentrate on creating and deploying AI applications, instead of waiting for infrastructure to catch up.
The model is already being put to work. Sharon AI intends to deploy up to 40,000 of the company’s Grace Blackwell GB300 GPUs, while Firmus Technologies is developing an AI factory campus in Batam, Indonesia, which is expected to eventually support up to 170,000 of the company’s GPUs across a 360-megawatt facility.
The effort is especially of interest to AI-native companies like Baseten, Fireworks AI and Together AI that require immediate access to vast amounts of computing power for model training, fine-tuning and high-volume AI inference, NVIDIA said.
And as demand for AI services continues to ramp up, those companies also need infrastructure that can scale quickly without requiring them to make massive upfront investments.
The announcement underlines NVIDIA’s broader roadmap as the AI market matures.
It is not only about becoming a chip supplier, but a long-term infrastructure partner for the AI ecosystem. NVIDIA sees the future of AI being built not just on powerful chips, but on accessible, scalable computing infrastructure, and is betting that the way to get there is by helping cloud providers finance and deploy AI factories while sharing in the revenue they generate.
