NextFin News - Firmus Technologies has turned an AI infrastructure pitch into a concrete regional buildout, saying it will deploy 170,000 Nvidia graphics processing units in Batam, Indonesia, from the first quarter of 2027 to the start of 2028. The deal is not just a supply agreement. Firmus said the partnership will let it buy Nvidia infrastructure and sell Nvidia-powered cloud services to "AI Native" customers and others, while Nvidia earns product revenue and a share of cloud revenue.
The scale is what makes the announcement consequential. Firmus said it expects the arrangement to generate up to $30 billion in revenue over the first six years, based on customer commitments. That figure is not booked revenue, but it still signals how aggressively the company is trying to position itself as a regional AI infrastructure platform rather than a niche startup. A deployment of 170,000 GPUs is large by any regional standard and indicates a project built around sustained demand, not a one-off pilot.
Batam is the geographic clue that gives the deal its strategic edge. The Indonesian island sits close to Singapore, which has become one of Asia’s most important cloud and finance hubs. For data-center operators, that location can offer a practical compromise: close enough to serve regional customers with manageable latency, but with more room for land, power and construction than the city-state itself. In AI infrastructure, those physical constraints increasingly matter as much as the chips.
The announcement also shows how Nvidia’s role in the AI economy keeps broadening. Firmus said the U.S.-listed chip maker will receive product revenue and a share of cloud revenue, linking Nvidia directly to the economics of deployment rather than only to hardware sales. That structure suggests the company is trying to capture more of the value chain as operators look for partners that can help finance and scale large compute projects.
Firmus has been building toward this moment for months. In April, it said it had raised $1.35 billion over the previous six months, which gave it a $5.5 billion post-money valuation. Reuters also reported that it had appointed investment banks to work on a possible initial public offering. Those financing signals matter because a project of this size requires not only demand, but sustained access to capital and a credible path to monetization.
The broader message is clear: AI infrastructure is moving deeper into Southeast Asia, and the race is no longer only about access to chips. It is about who can secure power, land, connectivity, financing and customers at the same time. Batam gives Firmus a location advantage, but the real test will be whether it can turn announced capacity into high utilization once the GPUs begin arriving in 2027.
Why The Batam Buildout Matters
The Batam location matters because it helps explain why this deal could be commercially viable. Large-scale AI infrastructure is constrained by physical inputs that are increasingly hard to assemble in prime urban markets. Power availability, cooling, interconnection and construction timelines can decide whether a project becomes revenue-generating infrastructure or remains a slide deck. Batam’s proximity to Singapore gives Firmus access to regional demand while avoiding some of the cost and land pressure associated with the smaller hub.
That geography also reflects a broader trend in Asia’s digital infrastructure market. As AI demand grows, operators are likely to distribute capacity across multiple jurisdictions rather than concentrate everything in one place. That can broaden supply and diversify risk, but it also makes execution harder because every site must clear the same hurdles around power, network quality and financing. The winners are likely to be the firms that can bundle those elements into a repeatable operating model.
Firmus is trying to do exactly that. By pairing a specific site with a large GPU order and a cloud-service model, the company is presenting the data center as a managed AI utility rather than a simple warehouse for servers. That distinction matters because the economics of AI infrastructure are shifting from one-time hardware purchases toward recurring service revenue, where utilization and customer retention matter as much as capex.
“We have worked to figure out how to close the gap between the cost benefits that the large guys have access to, which they do because they have great credit ratings, and the guys that are up and comers,” Firmus co-chief executive Tim Rosenfield said.
“This is actually a really material way to level the playing field a little bit to give the next a chance to compete with the big guys,” Rosenfield said.
Those remarks frame the company’s pitch: smaller AI firms often lack the financing terms and scale advantages available to the biggest players, so the infrastructure has to be packaged in a way that lowers the barrier to entry. If Firmus can do that at Batam, it could tap a customer base that wants access to Nvidia-class compute without building and financing its own data center stack.
What Nvidia Gets From The Structure
Nvidia’s participation is notable because it goes beyond a standard equipment sale. Firmus said the chip maker will earn product revenue and a share of cloud revenue, giving it an ongoing financial interest in how the capacity is used. That suggests Nvidia is comfortable extending its economic footprint deeper into the infrastructure layer when it believes the project can scale.
The arrangement also fits a larger pattern in the AI market: the companies that control scarce infrastructure are increasingly trying to move up and down the stack at the same time. Chip makers want more exposure to recurring revenue. Infrastructure operators want better access to hardware, financing and software support. Customers want flexibility and lower upfront costs. Partnerships like the one between Firmus and Nvidia are designed to meet all three goals at once.
That structure can be powerful when demand is strong. But it also raises the bar on execution. If the capacity is delivered on time and customers use it heavily, both sides benefit. If buildout slips, power costs rise or utilization lags, the economics weaken quickly because data centers are capital intensive and revenue only scales if the machines are busy.
Firmus said Nvidia has previously participated in its capital raisings, making it an investor in the Australian firm. That background matters because it suggests the partnership is not purely transactional. Nvidia is not just supplying a project; it is extending a relationship with a company that has already sought its support in the capital stack.
The Capital Markets Test
The deal also gives a glimpse of how private capital is financing the next wave of AI infrastructure. Firmus said in April that it had raised $1.35 billion over the prior six months and reached a $5.5 billion post-money valuation. Those numbers show investors were already willing to back a rapid scale-up before the Batam project was fully in the open.
That matters because infrastructure scale is expensive long before it becomes profitable. Data-center operators have to secure land, permits, power agreements, cooling systems and networking equipment well before they can collect meaningful recurring revenue. A startup with a multi-billion-dollar valuation can move faster, but it also has more pressure to justify that valuation with tangible progress.
A possible initial public offering would sharpen that test further. Public-market investors generally want more disclosure on contract durability, utilization and margin economics than private investors do. If Firmus wants to turn the Batam buildout into a lasting franchise, it will need to show that customer commitments translate into durable revenue rather than a one-off burst of demand.
The company’s own revenue guidance should be read in that light. Up to $30 billion over six years sounds large because it is large, but it remains contingent on customer commitments, pricing, delivery schedules and execution. For now, it is a ceiling that reveals the ambition of the project more than a guarantee of future cash flow.
What Comes Next
The immediate questions are operational. Can Firmus secure the power, construction and network capacity needed to deliver the GPUs on schedule? Can it convert customer commitments into long-term service contracts? Can it keep the project on track through a multi-year buildout in a market where everyone wants the same constrained inputs?
Those questions will decide whether the Batam project becomes a model for AI infrastructure expansion in Southeast Asia or another reminder that announcements are easier than execution. The timeline matters too: the first quarter of 2027 to the start of 2028 is far enough away that the project must survive changes in demand, pricing and capital markets before the hardware even lands.
For now, the deal says the next stage of the AI race is no longer just about who can buy the most chips. It is about who can turn chips into a functioning platform, with the right geography, the right financing and the right customer base. Batam gives Firmus a starting point. The rest will depend on whether the company can make the economics work at scale.
The most important number in the announcement is not 170,000 GPUs or even the $30 billion revenue target. It is the distance between promised capacity and proven utilization. In AI infrastructure, that gap is where the real business is won or lost.
Explore more exclusive insights at nextfin.ai.
