NVIDIA has launched a new business model designed to accelerate access to high-performance computing for AI startups, model builders, and enterprises. By implementing a revenue-sharing and credit-support structure, the company enables AI cloud providers to procure infrastructure without the traditional capital-intensive barriers, according to the company’s official announcement.
How the NVIDIA AI Factory Model Works
The initiative shifts how AI infrastructure is financed and deployed. Instead of requiring upfront, long-term capital commitments, NVIDIA is partnering with AI cloud companies through a model that aligns economic interests. Under this framework, AI clouds purchase NVIDIA infrastructure, and the company earns both standard product revenue and a share of the resulting cloud service revenue.
This structure is intended to bring "AI factories"—data centers optimized for token-scale inference and model training—online more rapidly. By removing the wait times associated with traditional site selection, power procurement, and hardware integration, NVIDIA aims to help AI-native companies move from pilot programs to production at scale.
Early Adopters and Global Deployment
Several companies are already integrating this model into their infrastructure plans.
- Sharon AI: The company is currently deploying up to 40,000 NVIDIA Grace Blackwell GB300 GPUs. James Manning, cofounder and CEO of Sharon AI, stated that the collaboration is a "pivotal moment" in the company’s mission to provide sovereign, large-scale AI compute infrastructure.
- Firmus Technologies: The firm is constructing a DSX AI factory campus in Batam, Indonesia, designed to scale to 360 megawatts and house up to 170,000 NVIDIA GPUs. According to Tim Rosenfield, co-CEO of Firmus Technologies, this "DSX-aligned AI factory" is necessary for AI-native companies to remain competitive on a global scale.
Why Compute Access Matters for AI Natives
The demand for "token-scale" AI services has created a bottleneck for startups. Companies such as Baseten, Fireworks AI, and Together AI represent the growing segment of the market that requires immediate, scalable access to accelerated computing.
These organizations face a dual challenge: they need the raw power to handle high-volume agentic inference and model training, but they also require the commercial flexibility to scale their infrastructure as their product usage grows. This model addresses those needs by providing a direct pipeline to NVIDIA hardware, allowing these firms to bypass the logistical delays that have historically hindered high-growth AI organizations.
Frequently Asked Questions
What is an AI factory?
An AI factory is a data center architecture designed to serve customers and workloads across regions, optimized for continuous, high-volume AI inference and model training.
Who can access this new infrastructure?
The program is aimed at the fast-growing AI ecosystem of startups, model builders, enterprises, research organizations and regional AI players.
How does the revenue-sharing model differ from traditional procurement?
Traditional procurement requires significant upfront capital and long-term hardware commitments. The new model uses a revenue-sharing and credit-support system, which reduces the immediate financial burden on cloud providers and their customers.
Where can companies go to secure this capacity?
NVIDIA directs interested organizations to contact Sharon AI and Firmus to secure compute capacity and build and deploy AI models.