NVIDIA launches revenue-share compute financing model, signs Sharon AI and Firmus for 200,000+ GPUs

NVIDIA introduced a new business model on July 1, 2026, that lets AI cloud providers procure its infrastructure through revenue-sharing and credit-support arrangements rather than upfront capital alone, aiming to open large-scale compute access to startups and smaller AI clouds that have struggled to finance it.

What's new

NVIDIA's announcement describes the mechanics directly: "This new model enables AI clouds to procure NVIDIA infrastructure for AI-native, enterprise and ISV customers through economic alignment with a revenue-sharing and credit-support model." In practice, "AI clouds will sell NVIDIA-powered cloud services, with NVIDIA earning both standard product revenue and a share of the cloud revenue on the supported capacity" — turning what was historically a one-time hardware sale into a recurring, usage-linked revenue stream for NVIDIA while lowering the capital bar for cloud partners to stand up new capacity.

Two initial partners were named. Sharon AI is deploying up to 40,000 NVIDIA Grace Blackwell GB300 GPUs, with CEO James Manning calling it "a pivotal moment in Sharon AI's mission to deliver sovereign, large-scale AI compute infrastructure." Firmus is building a DSX AI factory campus in Batam, Indonesia, expected to scale to 360 megawatts and as many as 170,000 NVIDIA GPUs; co-CEO Tim Rosenfield said "AI-native companies need access to scalable, energy- and cost-efficient compute infrastructure to compete globally." NVIDIA cited AI-native inference companies including Baseten, Fireworks AI, and Together AI as examples of the demand driving the buildout.

Context

The announcement lands the same day NVIDIA and manufacturing partners detailed a separate $500 billion U.S. onshoring push spanning chip fabrication and systems assembly across 43 states — together forming a two-sided push to expand both where AI hardware is built and how the resulting compute gets financed and sold. NVIDIA has increasingly moved beyond selling GPUs outright toward structuring deals — investments, credit support, and now revenue-sharing — that tie its financial performance to how much of that hardware actually gets deployed and used, a pattern also visible in its recent large compute and cloud partnerships.

Why it matters

Access to financing, not chip supply alone, has been a binding constraint for smaller AI clouds and model builders trying to compete for GPU capacity against hyperscalers with deep balance sheets. By taking on credit support and sharing in downstream cloud revenue, NVIDIA effectively becomes a financial backer of the AI-cloud layer it sells hardware into — a structure that could accelerate capacity buildout at firms like Sharon AI and Firmus, but also deepens NVIDIA's economic exposure to how well that capacity gets utilized and sold. It's a sign the AI infrastructure boom is shifting from a pure hardware-purchasing cycle toward more elaborate financial engineering to keep new capacity coming online.