On July 1, according to The Information, Nvidia is quietly transforming from a pure chip seller into a kind of 'central bank' for the AI computing ecosystem. Its core move is to offer financial backstops to young GPU cloud providers: if these companies cannot find enough AI developers to rent their computing power, Nvidia will repurchase the unsold GPU capacity at a predetermined price. In return, Nvidia receives a percentage of these cloud providers' revenue, with the share gradually decreasing over the contract period.
New Business Model: Revenue Sharing + Credit Support
Nvidia officially announced the new model, which combines revenue sharing with credit support. It allows AI cloud providers to acquire Nvidia infrastructure without fully bearing the upfront capital expenditures, then offer compute services to AI-native companies, model developers, and enterprise clients. Internally, the project is referred to as the "AI Compute Partnership." A Nvidia spokesperson confirmed its existence.
Under the framework, Nvidia will share in cloud service revenues beyond standard product sales, creating a recurring revenue stream tied to usage. The core intent is to break the financing barrier that has long constrained startups from accessing large-scale AI computing. Nvidia positions this as the "DSX AI Factory" model, targeting AI service scenarios requiring cross-regional continuous operation, high utilization, and multi-tenant accelerated computing.
Early Participants: Sharon AI and Firmus
Sharon AI and Firmus are the first cloud providers to join. Sharon AI plans to deploy up to 40,000 Nvidia Grace Blackwell GB300 GPUs; Firmus is building a DSX AI Factory campus in Batam, Indonesia, expected to scale to 360 megawatts and host up to 170,000 Nvidia GPUs. These deployments showcase Nvidia's latest progress in converting compute demand into fundable, realizable infrastructure.
A data center executive commented that Nvidia's transaction "kills two birds with one stone." If Nvidia only backstopped the data center facility lease, the GPU financing problem would remain; but by promising to pay for unsold compute, both the GPU and data center financing issues are solved. Nvidia's backstop effectively acts as a credit enhancement tool, enabling emerging cloud providers to leverage larger capital for faster data center construction.
Strategic Context: Reducing Reliance on Hyperscalers
Currently, hyperscalers like Amazon, Microsoft, Oracle, Meta, and Google purchase the majority of Nvidia's chip output. Many of them are developing their own competitive AI chips. To mitigate this risk, Nvidia has been nurturing emerging GPU cloud providers such as CoreWeave for years. The AI Compute Partnership continues and deepens this strategy.
According to The Information, Nvidia has also recently discussed providing financial guarantees for OpenAI to lease a large data center in Ohio, which could cost up to $500 billion if fully built at current prices. Nvidia has already invested billions of dollars in several emerging cloud providers in exchange for equity, and in some cases agreed to repurchase chips from them, including CoreWeave and Lambda. Nvidia's own researchers have used GPU servers repurchased from Lambda.
Financial Commitments and Market Impact
In September 2024, Nvidia committed to buying all of CoreWeave's unsold capacity through 2032 if no tenants could be found, a contract valued at $6.3 billion. That move helped boost CoreWeave's stock nearly 30% in the following week. According to a regulatory filing in May, Nvidia added another $3.5 billion in guarantees for client data center leases in exchange for rights to purchase their stock.
Nvidia is building a multi-layered alignment of interests: equity investments, capacity repurchase, lease guarantees, and now revenue sharing. Each layer deepens the financial ties with downstream cloud providers, enabling Nvidia to directly capture incremental revenue from AI compute commercialization beyond chip sales. This shift marks Nvidia's evolution from a hardware vendor to a computational finance infrastructure provider.