NVIDIA will combine revenue sharing with credit support to finance AI cloud buildouts, a move announced July 1 in a blog post co-authored by CFO Colette Kress. The model targets the growing gap between AI startups' need for production-scale GPU infrastructure and their ability to secure traditional financing.
The shift to inference factories
AI adoption is moving from the era of training large language models to operating continuous production services. Inference workloads now demand AI factories that generate tokens at scale, requiring rapidly deployable GPUs, high utilization rates, and multi-tenant architectures. This shift puts pressure on AI cloud providers to expand capacity well before revenue from those workloads catches up.
For emerging AI companies, the capital requirements often outpace available funding. Large-scale GPU deployments can cost hundreds of millions of dollars, and traditional lenders have been slow to underwrite assets whose value depends on fast-moving AI demand curves.
A new financing framework
NVIDIA's model allows AI cloud providers to acquire infrastructure under a framework where NVIDIA shares in cloud revenue tied to the supported capacity. The company earns from both the initial hardware sale and a recurring, usage-based stream. "This approach is expected to accelerate adoption of NVIDIA's AI platform while creating a recurring, consumption-linked revenue stream," the company said.
For enterprises, AI startups, model developers, and ISVs, the result is faster access to compute. Customers can tap into existing GPU capacity instead of waiting for new data center construction, power provisioning, and hardware deployment. NVIDIA is effectively lowering the working-capital hurdle for cloud providers, which in turn shortens the time to production for end users.
For finance leaders monitoring infrastructure allocation, understanding revenue-sharing and credit-support mechanisms is becoming a core competency. Resources like the AI Learning Path for CFOs can help evaluate how these models affect capital planning and OpEx. Similarly, the growing intersection of AI infrastructure and financial strategy is a theme covered in AI for Finance.
First participants and demand drivers
NVIDIA identified Sharon AI and Firmus as early participants. Sharon AI plans to deploy up to 40,000 NVIDIA Grace Blackwell GB300 GPUs. Firmus is developing a DSX AI factory campus in Batam, Indonesia, with a planned scale of 360MW supporting up to 170,000 GPUs. Both providers will build DSX AI factories to serve regional and enterprise AI workloads.
On the demand side, AI-native cloud platforms such as Baseten, Fireworks AI, and Together AI are driving rapid consumption of on-demand accelerated computing. These platforms support workloads spanning model training, fine-tuning, post-training optimization, and large-scale inference.
Why this matters for finance professionals
The model turns a one-time hardware sale into a recurring revenue stream for NVIDIA while shifting infrastructure risk toward cloud providers. For CFOs and treasury teams, it signals a new form of vendor financing that blends credit support with variable, usage-based repayment. Monitoring the terms of these deals-and how they affect provider balance sheets-will be essential as AI factories become a larger share of corporate capital expenditure.
Your membership also unlocks:
