From Bitcoin to AI Factories: Firmus Just Jumped to a $10.5 Billion Test

The $2 billion raise put Firmus in a different valuation league

The debate around Firmus is no longer whether it can pivot. It is whether the market has already priced the story too far ahead of execution. A $2 billion equity raise pushed the company to a post-money valuation above $10.5 billion, while the latest funding brought total equity raised in the past year to more than $3 billion.

From Bitcoin mining to Nvidia DSX-powered AI infrastructure

Firmus is now positioned around Nvidia DSX-powered AI data centers and factories. The new capital is meant to accelerate Project Southgate in Australia and fund early development work in Indonesia and other Asia-Pacific markets. That changes the setup dramatically: the company now needs to turn power access, land, hardware procurement, and customer demand into live compute capacity on a fast schedule.

Why the backing matters

The investor mix also matters. Coatue and Nvidia participated again, while funds managed by Blackstone and Jane Street joined the round. That does not guarantee execution, but it does signal that investors are willing to fund scale at a large base valuation. For bulls, that strengthens the case that Firmus can move from power control to AI capacity quickly. For bears, it raises the cost of delays.

Firmus is trying to monetize power, design, and Nvidia-aligned supply chains

The core valuation question is straightforward: can Firmus convert megawatts and real estate into contracted compute supply fast enough to justify the valuation?

Immersion cooling and DSX design are meant to raise density

Firmus is building around immersion-cooled facility designs and liquid-cooled, high-density platforms intended to turn raw megawatts into usable training and inference capacity. In practical terms, the value is not just the building itself. It is how many AI workloads each megawatt can support and how quickly those workloads become revenue.

That shifts the valuation logic away from simple real-estate economics. A standard data-center lease is one thing; long-duration compute supply backed by liquid-cooled architecture and Nvidia's full-stack DSX blueprints could carry more weight if supply remains tight.

The Indonesia project makes the scale concrete

The pipeline is where the ambition becomes visible. In Indonesia, the local market already has roughly 580MW operational data-center capacity, with about 1.3GW more in projects. Firmus is not walking into an empty market; it is entering a region where AI infrastructure demand is accelerating.

At the center of that push is the Batam campus: a 360MW Nvidia DSX AI Factory campus developed with DayOne. At that scale, the project looks less like experimental speculation and more like a serious attempt to create large-format AI capacity in the Asia-Pacific region.

Customer demand, not just capacity, is the real test

What matters most is not just how much power Firmus can assemble, but who is paying for the output. In Indonesia, the Batam project is targeted at AI-native customers, and Firmus says the partnership could generate US$25 billion to US$30 billion in committed off-take agreements over six years. The arrangement also covers as many as 170,000 Nvidia AI accelerator chips through 2027 and 2028.

That is the bridge investors need to see actually hold together: secure power, lock in suitable sites, use liquid-cooled Nvidia-ready design, and convert customer interest into firm commitments before capacity scales.

The $30 billion claim is the bull case; execution is the fault line

After a funding round that dramatically raised expectations, the real question is whether ambition is turning into bankable supply. The bull case rests on a forward claim of as much as $30 billion in off-take agreements, tied to a campus already under construction and chip deliveries scheduled from the first quarter of 2027 through early 2028. If that pipeline converts into operating capacity and realized revenue, investors will have reason to treat Firmus as more than a development story.

Where the thesis could weaken

The supportive case is clean: Firmus is building around Nvidia's full-stack DSX blueprints and has access to as many as 170,000 Nvidia AI accelerator chips through the partnership. In a market short on liquid-cooled, Nvidia-aligned capacity, that positioning matters.

The weaker part of the story is still validation. The current thesis leans on committed off-take agreements rather than named tenants or traditional take-or-pay leases. It also depends on timelines holding and the revenue-sharing and credit support arrangement with Nvidia and DayOne translating into visible customer uptake.

If off-take remains unnamed, Batam slips, or Nvidia's involvement does not translate into visible customer commitments as capacity scales, the thesis weakens materially.