Nvidia launches revenue-sharing model with AI cloud startups

Nvidia ↗ 0.97% has unveiled a new business model that allows AI cloud providers to deploy its GPU infrastructure through revenue-sharing and credit-support arrangements, marking a shift from the chipmaker's traditional hardware sales approach toward recurring income tied to downstream cloud services.

Under the program, announced this week, AI cloud providers purchase Nvidia infrastructure without bearing full upfront capital costs. In return, Nvidia earns both standard product revenue from hardware sales and a share of cloud revenue generated from the supported capacity. If a cloud provider cannot find enough customers to rent its computing capacity, Nvidia will repurchase unsold GPU capacity at a predetermined price, serving as a financial backstop.aiweekly+1

First Partners Scale Up

Sharon AI and Firmus Technologies are among the first companies participating in the program. Sharon AI, listed on Nasdaq, signed a six-year collaboration with Nvidia worth up to $4.88 billion, deploying up to 40,000 Grace Blackwell GB300 GPUs across 72 megawatts of new AI data center capacity in Australia.sharonai+1

Firmus Technologies, an Australian AI infrastructure company, announced an eight-year partnership with Nvidia anchored by a 360-megawatt Nvidia DSX AI Factory campus in Batam, Indonesia. The deal covers up to 170,000 Nvidia AI accelerators across Grace-Blackwell, Vera-Rubin, and Vera platforms through 2027 and 2028. Firmus expects $25 billion to $30 billion in committed offtake agreements during the first six years of the partnership, according to the Straits Times.technode+2

From Chip Seller to Computing Ecosystem

The program represents Nvidia's broader evolution from a pure hardware vendor into what some observers have described as the "central bank" of the AI computing ecosystem. Nvidia already holds equity stakes in some partners, including Firmus, giving it layered financial exposure through hardware margins, equity appreciation, and now recurring revenue shares.kucoin+1

Nvidia also named Baseten, Fireworks AI, and Together AI as inference-side customers within the broader initiative. The company said the model is designed to serve a growing ecosystem of AI startups, model developers, enterprises, and regional AI entities that need access to large-scale compute but lack the capital to fund it independently.reddit+2

The combined first cohort of Sharon AI and Firmus brings roughly 432 megawatts of new committed AI data center capacity online under the revenue-sharing framework, with facilities expected to begin operations starting in early 2027.aiweekly+1