NVIDIA's New Business Model: Financial Guarantees for Revenue Sharing
On July 1, NVIDIA announced a new business model combining revenue sharing with credit support, dubbed internally the 'AI Compute Partnership.' Under this framework, NVIDIA provides a financial backstop to young cloud service providers that lease out its GPUs: if these companies fail to find enough AI developers to rent compute capacity, NVIDIA will repurchase the unsold GPU capacity at a pre-agreed price. In return, NVIDIA receives a percentage of the cloud providers' revenue, with the share gradually decreasing over the contract term. The model is designed to address a critical bottleneck: startup AI companies historically face severe restrictions in accessing capital-intensive infrastructure. Even long-term commitment contracts often fall short of securing financing for compute procurement. By realigning economic incentives, NVIDIA aims to enable faster deployment of full-stack accelerated computing without waiting for traditional infrastructure construction cycles.
Initial Deployments and Scale
Sharon AI and Firmus are the first participants. Sharon AI plans to deploy up to 40,000 NVIDIA Grace Blackwell GB300 GPUs, while Firmus is building a DSX AI factory campus on Batam Island, Indonesia, expected to scale to 360 MW with up to 170,000 NVIDIA GPUs. These deployments illustrate NVIDIA's progress in converting compute demand into financeable infrastructure. One data center executive noted that NVIDIA's approach 'kills two birds with one stone': solving both GPU financing and data center financing. By acting as a credit enhancement tool, NVIDIA enables emerging cloud providers—often with low credit ratings—to secure larger loans and accelerate facility construction.
Strategic Context: Reducing Dependence on Incumbent Cloud Giants
NVIDIA's strategy is driven by a clear competitive landscape. Today, a handful of large cloud providers—Amazon, Microsoft, SpaceX, Oracle, Meta, and Google—purchase the majority of NVIDIA's chip supply. However, several of these companies are developing their own competing AI chips, posing a long-term threat to NVIDIA. To mitigate this risk, NVIDIA has been cultivating a cohort of emerging GPU cloud providers, such as CoreWeave, for several years. The AI Compute Partnership deepens this strategy. According to The Information, NVIDIA is also in talks to provide financial guarantees for OpenAI to lease a large data center in Ohio, which could cost up to $500 billion to fully build out at current prices.
Capital Commitments and Capacity Guarantee Cases
NVIDIA's financial commitments in this direction are substantial. In September 2024, NVIDIA agreed to buy all of CoreWeave's unsold capacity through 2032 if it cannot find renters, in a deal valued at $6.3 billion. This commitment alleviated investor concerns about CoreWeave's high leverage, boosting its stock price by nearly 30% in the following week. According to a regulatory filing in May covering the quarter through April, NVIDIA added $3.5 billion in guarantees for customer data center leases in exchange for stock purchase rights. Overall, NVIDIA is building a multi-layered binding mechanism: equity investments, capacity buybacks, lease guarantees, and now revenue sharing. Each layer deepens NVIDIA's financial ties with downstream cloud providers, allowing it to capture incremental revenue from AI compute commercialization beyond chip sales.
