Nvidia Launches Revenue-Sharing Model to Ease AI Infrastructure Costs for Cloud Providers

Introduction

NVIDIA has introduced a new financing arrangement designed to lower the upfront cost of deploying its AI computing hardware, shifting a portion of its revenue model toward long-term, usage-based earnings tied to the cloud services built on its chips.

The program was detailed this week in a blog post co-authored by Nvidia Chief Financial Officer Colette Kress. NVIDIA unveiled the optional revenue-sharing and credit-support financing model on July 1, 2026. Under the arrangement, participating AI cloud operators buy and deploy Nvidia’s infrastructure, then sell cloud services built on that hardware to model developers, inference providers, agent platforms, and enterprise customers. NVIDIA earns its standard hardware margin on the initial sale and then takes an additional share of the cloud revenue generated on the supported capacity, creating a usage-linked earnings stream that extends beyond the point of sale.

Nvidia framed the move as a direct response to how enterprise AI demand is evolving. The company said the shift reflects a broader move from model development toward production inference, where demand is increasingly for continuously operating GPU data centers optimized for always-on training and inference at scale. For customers, the arrangement can mean getting full-stack accelerated computing sooner, without having to wait for site selection, power procurement, construction and hardware bring-up.

A Shift in How Nvidia Gets Paid

The financing model marks a departure from Nvidia’s traditional sales approach, in which customers pay upfront for hardware regardless of how it is later used. Analysts covering the announcement described it as a structural change in how the company monetizes its dominance in AI chips. The structure lets participating AI clouds draw token credits against future capacity now, while Nvidia collects standard hardware revenue plus a recurring cut of the cloud income that capacity generates, shifting Nvidia from a pure equipment vendor toward a financier with a stake in customer utilization.

That shift comes with new complexity for the cloud operators who sign on. Platform teams now have to account for the fact that while capital expenditure drops under the new model, long-term revenue-share obligations and utilization tracking become an ongoing part of the deal.

Some reporting has also pointed to additional guarantees built into the program beyond simple revenue-sharing. According to coverage citing sourcing from The Information, the framework guarantees GPU capacity for startup cloud providers, with Nvidia taking a revenue share or equity stake in return for backstopping idle hardware. Under that arrangement, if a partner’s GPUs sit unused, Nvidia will fund the rental of that idle capacity or buy it back outright, effectively acting as a credit endorsement that helps smaller cloud providers secure the large-scale financing needed for data center construction. That kind of backstop extends a model Nvidia has already established with CoreWeave, to which it has committed a $6.3 billion guarantee for unsold computing capacity through 2032. The same reporting indicated Nvidia is also negotiating a similar financial backstop tied to OpenAI’s planned data center projects, though Nvidia has not confirmed that arrangement publicly.

First Partners: Sharon AI and Firmus

Nvidia named two companies as its first partners under the new model: Sharon AI, an Australian sovereign-cloud provider, and Firmus Technologies, which is co-headquartered in Singapore.

Sharon AI is deploying up to 40,000 Nvidia Grace Blackwell GB300 GPUs, while Firmus is developing a DSX AI factory campus in Batam, Indonesia, that is expected to scale to 360 megawatts and support up to 170,000 Nvidia GPUs. Together, the two projects represent one of the largest publicly disclosed non-hyperscaler GPU commitments to date. Combined, the deal could scale to roughly 210,000 GPUs across both partners.

Sharon AI’s growth plans extend beyond the initial 40,000-unit deployment. The company said in a June 12 release that it expects to have more than 55,000 total Nvidia GPUs deployed by mid-2027. Additional reporting on the scale of the two companies’ broader operations put the combined power commitment even higher: Firmus and Sharon AI are together committing 432 megawatts across Indonesia and Australia, with Sharon AI already having 102 megawatts contracted to end customers and Firmus projecting between $25 billion and $30 billion in offtake over six years.

The Batam campus, in particular, is being built through a broader partnership network. Firmus’s Batam campus is being developed with DayOne, a Singapore-headquartered digital infrastructure platform, under an eight-year partnership with Nvidia extending to 2034. The facility is expected to incorporate Nvidia’s Grace-Blackwell, Vera-Rubin and Vera chip platforms as they roll out through 2027 and 2028.

Nvidia also gestured toward a wider circle of companies benefiting from the new financing structure on the demand side. Alongside Sharon AI and Firmus as the named infrastructure partners, Nvidia gave a broader shout-out to Baseten, Fireworks AI and Together AI as inference-side customers expected to make use of the expanded capacity.

Executives React

Leaders at both launch partners framed the announcement as a turning point for their companies’ growth plans.

James Manning, co-founder and chief executive of Sharon AI, said the arrangement with Nvidia “marks a pivotal moment” in the company’s effort to build out sovereign, large-scale AI compute capacity for its home market. Firmus co-CEO Tim Rosenfield struck a similar note, arguing that AI-native companies need infrastructure that is both scalable and cost-efficient in order to compete on a global stage, and said the company’s Nvidia-aligned facility would help organizations get faster access to the computing power needed to build and scale AI applications.

Why Nvidia Is Making the Change

Industry observers have pointed to a persistent bottleneck in AI infrastructure as the underlying driver of Nvidia’s new approach. Access to GPU capacity has remained one of the most significant constraints on AI companies’ growth, even as customer demand for AI services continues to climb. Startups and regional cloud operators in particular have often struggled to raise the capital needed to build out data centers, despite having customers lined up and ready to buy compute.

By tying its own revenue to the performance of the infrastructure it finances, Nvidia is betting that lowering the barrier to entry for smaller cloud operators will expand the overall market for its chips faster than requiring full upfront payment would. The initiative also diversifies Nvidia’s customer base beyond hyperscalers such as Microsoft Azure, AWS and Google Cloud, extending its reach toward a wider set of independent AI infrastructure developers.

Nvidia has also framed the initiative within its existing data center strategy. The company said several cloud operators are already building AI facilities using its DSX data center platform, and positioned the new financing structure as a way to make that technology more attainable for emerging firms facing steep infrastructure costs.

Market Reaction

The announcement landed as Nvidia continues to trade near record valuations. Nvidia shares traded at $195 on the day of the announcement, giving the company a market capitalization of approximately $4.7 trillion. That scale underscores the stakes of Nvidia’s shift toward a financing-and-revenue-share model: even a modest share of recurring cloud revenue across hundreds of thousands of deployed GPUs could represent a meaningful new income stream layered on top of hardware sales.

Coverage of the announcement across financial and technology media was broadly consistent, though outlets differed slightly in how they characterized the program’s formal name and scope some referred to it simply as a revenue-sharing model, while others described it as part of a wider “AI Computing Partner Program” encompassing both financing and capacity guarantees. The core mechanics, however, were consistently described: reduced upfront costs for cloud operators, ongoing revenue participation for Nvidia, and an initial rollout centered on Sharon AI and Firmus.