NVIDIA has unveiled a new business model for AI cloud providers that combines revenue sharing with credit support, aiming to make large-scale AI infrastructure more accessible to startups, AI model developers, enterprises, research institutions, and regional cloud providers.
The initiative comes as AI workloads increasingly shift from training models to running inference in production, where AI factories operate around the clock to generate AI outputs at scale.
Under the new framework, AI cloud providers will purchase NVIDIA’s infrastructure and deliver cloud services powered by its technology. NVIDIA will generate revenue not only through hardware sales but also by receiving a share of the cloud revenue produced from the supported infrastructure.
The company said the model aligns its interests with AI cloud providers by offering a revenue-sharing and credit-support structure, helping them serve AI-focused businesses, enterprises, and independent software vendors more efficiently.
NVIDIA said the approach is designed to give AI companies faster access to computing resources without waiting for lengthy data centre development, including site selection, power availability, construction, and hardware installation.
The programme is already being rolled out, with several AI cloud providers developing NVIDIA DSX AI factories in different regions. Among the first adopters are Sharon AI and Firmus.
Sharon AI plans to deploy as many as 40,000 NVIDIA Grace Blackwell GB300 GPUs to expand its AI computing capabilities.
Firmus, meanwhile, is developing a DSX AI factory campus in Batam, Indonesia. The facility is expected to grow to 360 megawatts of capacity and eventually support up to 170,000 NVIDIA GPUs.
According to NVIDIA, the new model is intended to support AI-native companies, including foundation model developers, inference service providers, AI agent platforms, and enterprises that require large-scale computing for training, fine-tuning, post-training, and agentic AI inference. The company added that as AI deployments move beyond pilot projects into full-scale production, organisations are increasingly seeking both dependable computing capacity and greater commercial flexibility.
Also Read: NVIDIA Launches Open-Source AI Models for Physical World Applications
