OpenAI becomes anchor customer at Firmus, pushing the Nvidia-backed builder past 900MW
A multi-year deal for dedicated capacity at two new Malaysian AI factory sites is Firmus's first named hyperscaler customer -- and a bet that Southeast Asia becomes a compute exporter, not just a consumer.
Firmus, the Nvidia-backed AI infrastructure builder headquartered in Sydney, announced on September 8, 2026 that OpenAI has signed a multi-year agreement for dedicated AI compute capacity at two new Firmus AI Factory sites in Malaysia, making OpenAI the company's first named anchor customer there. The deal pushes Firmus's total contracted capacity across all customers past 900MW, spread across seven AI factory sites in Australia, Singapore, Indonesia and Malaysia -- two of them operational today, five still under development. The Malaysian sites will run Nvidia's Vera Rubin NVL72 systems on Nvidia's DSX AI Factory Platform, integrated with Firmus's own HyperCube infrastructure design.
A regional bet stated plainly
Firmus co-founder and co-CEO Tim Rosenfield framed the deal as more than a customer win, describing it as the moment "Asia-Pacific becomes a producer of intelligence, not just a consumer of it." OpenAI, for its part, described the region's AI infrastructure demand as accelerating fast enough to require factories built for it from the ground up, not retrofitted general-purpose data centers.
Why the concentration matters
Five of Firmus's seven sites are still under construction, with a target of bringing them into service over the next 24 months -- so most of the newly announced 900MW is contracted, not yet delivered. That gap between signed capacity and live capacity is the same pattern showing up across the AI infrastructure buildout broadly: financing and offtake commitments arrive well ahead of the buildings and power hookups behind them, which is worth tracking as its own risk category separate from the headline megawatt or dollar figures.