Firmus Technologies is making a big move on Batam.
The Australian AI infrastructure company has signed an eight-year partnership with Nvidia tied to a planned 360MW AI data center on the Indonesian island.
The project would use up to 170,000 Nvidia GPUs, with deliveries expected across 2027 and 2028, according to Bloomberg.
The plan puts Batam deeper into Southeast Asia’s race to host AI compute capacity. It also gives Nvidia another regional infrastructure partner at a time when demand for AI chips continues to stretch supply.
Batam buildout
Bloomberg reported that the Batam project is being built with Singapore-based DayOne, which is backed by investors including Coatue Management and SoftBank. DayOne has already been expanding its AI infrastructure footprint, including a $2 billion funding round tied to data center growth.
The Firmus project is large even by AI data center standards. A 360MW campus would require major power, cooling, and networking support, especially if the site is expected to host dense GPU clusters for AI training and inference.
The timeline is also still forward-looking. GPU deliveries are expected to span 2027 and 2028, with initial operations targeted for 2027. That means the Batam project is not capacity customers can use today. It is a planned regional buildout that still has to move through delivery, construction, power, and customer-readiness milestones.
That distinction is important for cloud buyers and AI teams. A 360MW facility near Singapore could eventually add meaningful compute capacity in Southeast Asia, but the public reporting does not yet confirm final power sourcing, named customer commitments, or full delivery timing.
AI capacity near Singapore
Batam’s location is part of the appeal. The island sits close to Singapore, where data center growth has been shaped by space, power, and sustainability constraints. For AI cloud providers, nearby capacity can create another route into regional demand without putting every new megawatt inside Singapore.
Power remains one of the biggest practical questions. AI data centers are expensive to build and difficult to bring online quickly, especially when GPU density pushes cooling and electricity needs higher.
TechRepublic’s coverage of power and cooling trends explains why electricity, thermal design, and infrastructure planning are becoming central to large AI deployments.
The Firmus plan also lands as AI infrastructure investment spreads across APAC. Recent regional moves include ByteDance’s Malaysia AI chip deal, another example of how chip access, location, and export controls are shaping where AI capacity gets built.
For now, the Batam project gives Firmus and Nvidia a large regional AI infrastructure play, but it should still be treated as a buildout in progress. If Firmus, Nvidia, and DayOne stay on schedule, the site could become a major AI compute hub near Singapore.
Until then, the important updates will be power agreements, customer names, GPU delivery progress, and the first operational milestone.
Also read: China’s LineShine tops the supercomputer list, adding another APAC marker in the race for advanced computing capacity.