Back
Nvidia advances AI compute partnership to rent startup GPU capacity

Nvidia advances AI compute partnership to rent startup GPU capacity

Nvidia is broadening its AI infrastructure model with revenue sharing and token credits for startups, using partner capacity in Australia and Indonesia to expand access to scarce computing power.

Fact Check
The official Nvidia blog announcement (July 1, 2026) and CNBC both confirm Nvidia is broadening its AI infrastructure model through revenue sharing and compute credits targeting startups and emerging AI companies. The blog explicitly describes a revenue-share and credit-support model tied to token-scale inference demand, and names partners Sharon AI (Australia) and Firmus (Batam, Indonesia). The claim's mention of Australia and Indonesia partner capacity matches Sharon AI's Australian data center footprint and Firmus's Indonesian AI factory. All core elements are corroborated by the primary source.
    Reference: 123
Summary

Nvidia is expanding its AI infrastructure partnership strategy with revenue-sharing agreements that let fast-growing startups exchange access to compute power for a share of future profits. The program offers token credits to support development by cloud-based AI firms, model builders and other enterprises, while positioning Nvidia as an intermediary connecting startups to full-stack computing built on its chips. Nvidia named two initial partners for the scheme: Australia-based Sharon AI, which plans to deploy up to 40,000 Nvidia GPUs, and Singapore AI infrastructure company Firmus Technologies, which is building a data center in Batam, Indonesia, expected to scale to 360 megawatts and house up to 170,000 Nvidia GPUs. The move underscores how scarce and expensive GPU access has become for AI startups, driving companies toward revenue-sharing and other alternative financing arrangements. It also adds to Nvidia's broader effort to deepen its role in AI infrastructure economics beyond traditional chip sales.

Terms & Concepts
  • GPU capacity: Available graphics processing power that companies can use for AI training, inference and other compute-intensive tasks.
  • full-stack computing: An integrated computing setup that combines hardware, software and infrastructure needed to build and run AI systems.
  • token credits: Usage credits that can be applied toward computing resources under Nvidia's partnership program.