Firmus is teaming up with Nvidia to build a major data center in Indonesia, according to The Information. The deal puts one of the most aggressive players in sustainable AI infrastructure together with the company whose chips power nearly every serious AI buildout on the planet. And it lands the project in Southeast Asia, the region everyone in this race is suddenly fighting over.
The Information reports the project as a large-scale facility, the kind built specifically to run dense clusters of Nvidia GPUs for AI training and inference. The source is light on hard numbers so far, but the signal is clear enough on its own.
Who Firmus is
If the name doesn’t ring a bell, here’s the short version. Firmus runs Sustainable Metal Cloud, an AI infrastructure platform known for immersion cooling, where servers sit in special fluid instead of relying on fans and air conditioning. That approach cuts power and water use, which matters a lot when you’re packing thousands of power-hungry chips into one building.
Firmus has worked closely with Nvidia before, running GPU clusters in Singapore and pitching itself as a greener way to scale AI compute. This Indonesia move extends that playbook into a new market.
Why this matters
What stands out here is the location. Most of the AI data center money has poured into the US, with a second wave hitting Europe and the Gulf. Southeast Asia is the next frontier, and Indonesia is courting these investments hard.
A few reasons the region is heating up:
- Power and land. Big AI campuses need enormous, steady electricity and lots of space. Established hubs like Singapore have capped new data center growth over energy concerns, pushing builders to look next door.
- Demand. Indonesia is the largest economy in Southeast Asia with a huge, young, online population. Local AI demand is real, not theoretical.
- Chip access. Getting Nvidia hardware into the region, legally and at scale, is its own competitive edge. A direct Nvidia partnership smooths that path.
For practitioners, this is another sign that AI compute is going global fast. The bottleneck isn’t just chips anymore. It’s power, cooling, and where you can physically put the machines.
How it compares
The status quo until recently was simple. You wanted serious GPU capacity, you rented it from a US hyperscaler or waited in line. That’s changing. Specialized players like Firmus, CoreWeave, and a growing list of regional operators are building dedicated AI capacity outside the big three clouds, often with Nvidia as a direct partner rather than a parts supplier.
Nvidia clearly likes this pattern. Spreading its chips across independent operators in new geographies widens its footprint and locks in demand, without leaning entirely on Amazon, Microsoft, and Google. Indonesia fits that strategy neatly.
The sustainability angle is worth flagging too. As AI’s energy appetite draws more scrutiny, immersion cooling and efficient design aren’t just nice marketing. They’re becoming a requirement to win government approval and grid access in places watching their power budgets.
What to watch next
A few things to keep an eye on as details firm up:
- Scale and timeline. How big is the facility, how many GPUs, and when does it come online?
- Power source. Indonesia leans on coal. Whether this runs on cleaner energy will shape how the sustainability claims hold up.
- Customers. Is this capacity for local enterprises and governments, regional AI startups, or international buyers renting from afar?
- Copycats. Expect more of these regional Nvidia partnerships to surface across Asia if this one moves quickly.
The race to build AI infrastructure is no longer just an American story. It’s spreading to wherever there’s power, space, and willing governments. Indonesia just got on the map.
Full details are available at The Information.