170,000 Nvidia GPUs. Nvidia on July 2, 2026 announced a partnership program that lets fast-growing AI startups swap access to compute power for a share of future product and cloud revenue. The company said the program will grant token credits to qualifying startups, positioning Nvidia as an intermediary between cloud operators and model builders while taking slices of downstream revenue rather than only cash up front. Nvidia presented the initiative as an extension of its DSX data centre platform and said the model is aimed at lowering the financial barrier to high-performance infrastructure.

170,000 Nvidia GPUs. That figure represents the potential scale of a Batam, Indonesia facility being built by Firmus Technologies and it helps explain the ambition behind the new commercial structure. Under the programme, cloud providers will sell Nvidia-powered cloud services and share both standard hardware sales revenue and a portion of their future cloud earnings with Nvidia. Nvidia described the offering as a revenue-sharing and credit-support model rather than a conventional hardware sale.

How the program works

Nvidia said the company will grant token credits to qualifying startups that can be used to access compute on cloud services powered by Nvidia processors. Cloud providers in turn will sell those Nvidia-powered services and share revenue back with Nvidia, giving the chipmaker both product revenue from deployments and a slice of the cloud earnings those deployments generate. The firm named Sharon AI of Australia as one of the initial partners and said Sharon AI will deploy up to 40,000 GPUs under the scheme.

Firmus Technologies, based in Singapore, is building the Batam data centre that Nvidia said could scale to 360 megawatts and house up to 170,000 Nvidia GPUs. The arrangement is pitched at cloud-based AI firms, model builders and other enterprises that require large volumes of GPU compute but face capital constraints or spot-market price volatility. Nvidia framed the program as a way to lower the financial barrier to high-performance infrastructure by providing compute credits in exchange for future revenue from products and cloud services that use Nvidia-powered infrastructure.

Market context and company strategy

Nvidia presented the move against a backdrop of constrained GPU supply and rising demand from generative AI workloads, a dynamic the company and industry commentary compared to a scarce commodity. The programme is part of a broader push to expand Nvidia's footprint in cloud infrastructure and data centres, and to deepen customer relationships by broadening adoption of its processors.

The company said the revenue-sharing option lets it act as an intermediary between cloud operators and model builders while taking stakes in downstream cloud earnings. Nvidia didn't publish detailed percentage splits or contract terms in the announcement, and it named Sharon AI and Firmus Technologies as the initial commercial partners behind the rollout.

A wider shift toward revenue-sharing or equity-sharing arrangements has emerged among AI firms and infrastructure providers as a way to secure access to costly compute. Nvidia said cloud operators are already building AI facilities on its DSX platform, and the revenue-sharing option gives Nvidia both product revenue from hardware deployments and a stake in the cloud earnings those deployments generate.

Related Articles

One concrete detail to watch: Sharon AI has said it may deploy up to 40,000 GPUs under the scheme, and Nvidia highlighted a Batam campus being built by Firmus Technologies that could scale to 170,000 GPUs and about 360 megawatts of capacity. Originally reported by CNBC.

This article was created with AI assistance.