Firmus, backed by Nvidia, signed a multi-year deal to supply OpenAI with computing capacity from two Malaysian data centers, pushing total contracted capacity for all customers above 900 megawatts.

Firmus plans seven AI factories across Australia, Singapore, Indonesia and Malaysia—two already operational and five more under development—with deliveries over the next 24 months, and media reports suggest discussions around an Australian Securities Exchange listing.

OpenAI’s VP of Compute Strategy says the new sites will help meet rising regional and global demand for OpenAI products.

The 900 megawatts figure reflects capacity, not guaranteed revenue, highlighting uncertainty about exact financial gains from the deal.

Nvidia’s stock dipped about 1.5% following the news, trading around $226.86 as markets opened.

The HyperCube architecture is designed to accelerate data-center deployment, boost capital efficiency and performance, and reduce AI token production costs.

Investors should note that the precise scale and revenue impact for Nvidia and OpenAI remain uncertain until procurement quantities and delivery timelines are disclosed.

All systems are manufactured and prefabricated in regional New South Wales to improve supply control and deployment speed.

The information comes from a Firmus media release reported by Hugo Mathers.

Firmus employs an energy-in, tokens-out model aimed at maximizing efficiency of energy use and infrastructure investment for AI compute output, leveraging proprietary cooling and energy tech.

Nvidia’s broader OpenAI collaboration envisions deploying Vera Rubin-based infrastructure with scaling targets, while the deal with Firmus adds regional capacity via APAC axes.

Anchoring customers like OpenAI could help finance Firmus’ infrastructure build-out, potentially accelerating Malaysian facilities and future Vera Rubin purchases.