With the rapidly growing demand for infrastructure to run advanced AI models, Australian AI infrastructure company Firmus Technologies announced on Tuesday (September 8th) that it has established a multi-year strategic partnership with OpenAI. Under the agreement, OpenAI will lease the dedicated computing capacity of Firmus’s 2 AI factories in Malaysia and become the main client at these two sites.
Despite not disclosing the specific financial details of the collaboration, the partnership expands Firmus’s business footprint in the Asia-Pacific region, boosting the contracted capacity for all clients to over 900 megawatts (MW).
Co-founder and co-CEO of Firmus, Tim Rosenfield, stated, “This multi-year partnership signifies the transition in the Asia-Pacific region from AI consumers to AI producers.” He mentioned that Firmus has spent 7 years establishing a business layout from Southeast Asia to Australia, combining manufacturing, technology, and teams to rapidly implement business initiatives.
OpenAI sees this collaboration as an expansion of its global computing capabilities. Sachin Katti, Vice President of Computing Strategy at the company, expressed, “We are delighted to collaborate with Firmus to jointly expand our global computing capabilities.” He added, “These AI factories in Malaysia will help meet the rapidly growing demand for OpenAI products in the region. We look forward to enabling these computing capabilities together.”
Currently, Firmus is developing 7 AI factories in Australia, Singapore, Indonesia, and Malaysia. Two of them in Singapore and Australia have been officially operational, with the remaining 5 expected to be completed in about 2 years.
As part of the collaboration, Firmus will deploy NVIDIA’s new generation Vera Rubin acceleration computing platform on a large scale in the Asia-Pacific region. The inclusion of the Malaysia site further expands its footprint in the region.
However, with Malaysia being seen as the fastest-growing AI data center market in Southeast Asia, its rapid expansion has raised concerns in the industry about water and electricity consumption, potentially adding pressure or instability to the local supply.
These AI factories will utilize NVIDIA’s DSX AI Factory platform with Vera Rubin NVL72 rack-level systems, deeply integrated with Firmus’s in-house developed HyperCube liquid cooling systems, mechanical systems, and electrification technologies.
Firmus stated that integrating more parts of the AI factory value chain together can increase deployment speed, capital efficiency, operational performance, and ultimately reduce the production costs of AI tokens.
Nico Caprez, Vice President of Global AI Infrastructure Growth at NVIDIA, stated that the demand for AI infrastructure in the Asia-Pacific region is accelerating. To meet this demand, AI factories need to possess characteristics such as high performance, durability, efficiency, scalability, and interchangeable workloads.
In June 2026, Firmus announced a strategic computing partnership with NVIDIA, with the collaboration scheduled until 2034. The core of the partnership involves building a 360-megawatt NVIDIA DSX AI factory in Batam, Indonesia, where as many as 170,000 NVIDIA AI accelerators will be deployed.
At the time, Firmus stated that within the first six years of the collaboration, they expect to generate revenue of $25 billion to $30 billion from the committed purchase agreements.
This agreement was reached before rumors circulated in the market that Firmus would conduct its initial public offering (IPO) this year, which is expected to be one of the largest listings in Australia in recent years but has not been officially confirmed.
Over the past year, the company has undergone several rounds of equity financing, raising over $30 billion. The market estimates the company’s overall valuation to exceed $10.5 billion, with backers including NVIDIA, Jane Street Capital, Blackstone Group funds, and Coatue Management.
Furthermore, in February of this year, Blackstone also provided an additional $10 billion in debt financing for the company.