OpenAI has signed a multi-year deal to take AI compute from two new Firmus sites in Malaysia, making the ChatGPT maker an anchor customer of the Nvidia-backed data center startup.
Firmus co-chief executive Tim Rosenfield called the pact the moment Asia-Pacific becomes a producer of intelligence rather than just a consumer of it. OpenAI’s vice president of compute strategy, Sachin Katti, said the Malaysian capacity will help serve growing demand across the region and around the world.
The agreement pushes Firmus’s total contracted capacity beyond 900 megawatts. The startup now counts seven AI factories across Australia, Singapore, Indonesia and Malaysia, though only two are operating today. The remaining sites are expected to be complete within 24 months.
Firmus has spent seven years building a footprint from Southeast Asia to Australia, Rosenfield said, including the manufacturing and engineering teams needed to bring sites online quickly. The company banked $2B in August, one of the larger raises of the AI infrastructure boom, to accelerate that buildout.
The deal also hands Nvidia another win in the race to place its accelerators: the chip giant backs Firmus, and long-term commitments from anchor tenants like OpenAI lock in years of demand for the factories it helps finance. Malaysia has become a favored location for such projects thanks to available land and power.
