Australian AI infrastructure company Firmus Technologies has expanded its relationship with Meta into Southeast Asia, signing agreements for GPU compute capacity at AI factories being developed in the region. The agreements cover contracted capacity as well as options for further expansion, with Meta planning to use the infrastructure for AI research, model development and training.
The deal extends a relationship that already includes Firmus’s Australian operations. Meta is currently using Nvidia GB300 NVL72 systems at Firmus’s Melbourne AI factory, which Firmus claims is the largest deployment of Nvidia Blackwell Ultra systems in the Southern Hemisphere.
Firmus did not disclose the amount of capacity covered by the new agreements or their financial value.
The Southeast Asian facilities will be built using Nvidia’s DSX AI Factory platform, with Firmus integrating the computing infrastructure with its own HyperCube platform. HyperCube combines liquid cooling, power and mechanical systems into factory-built modules that Firmus manufactures in regional New South Wales. The latest generation supports Nvidia’s GB300 and Vera Rubin NVL72 systems, as well as Spectrum-X networking.
Firmus says the approach is intended to integrate computing, networking, cooling and power at the infrastructure level, allowing AI factories to be deployed more rapidly and operate with greater energy efficiency.
“We see Firmus as a long-term strategic infrastructure partner as we build for the next generation of AI across Asia-Pacific and expand the infrastructure supporting our AI research and model development efforts – the foundation of personal superintelligence,” said Gaya Nagarajan, vice president of engineering, infrastructure at Meta.
Tim Rosenfield, co-founder and co-CEO of Firmus, said the infrastructure was being designed for customers whose workloads require high-performance computing and energy efficiency. “Meta’s AI research and model development activities require infrastructure capable of supporting some of the world’s most demanding AI workloads. Firmus’s AI Factory build-out across Asia-Pacific has the scale to support and grow these ambitions,” he said.
Growing customer base
The agreement adds another major technology company to Firmus’s growing customer base as it expands its AI Factory platform beyond Australia. Firmus announced a multi-year agreement with OpenAI earlier this month for dedicated AI compute capacity at two Malaysian AI factories. That agreement made OpenAI an anchor customer and took Firmus’s total contracted capacity across its customer base above 900MW.
Firmus said at the time that its portfolio comprised seven AI factories across Australia, Singapore, Indonesia and Malaysia, with two operational sites and five under development. The company’s Australian expansion is also continuing. Firmus is developing AI infrastructure projects in Tasmania and South Australia, while its manufacturing operation in regional NSW supplies the HyperCube infrastructure used by its AI Factory platform.
The company’s Australian strategy has increasingly focused on integrating power procurement, generation, cooling and compute as opposed to treating the data centre as a conventional powered shell. In a recent submission to a Tasmanian government inquiry, Firmus argued that large AI infrastructure projects should contribute to the generation and grid infrastructure they require, while supporting flexible operation of their electricity demand.