Firmus' $10B Backing Won't Save You if the IPO Prices in a $6B Dream

Firmus is already a mature valuation story, not an early startup story

Firmus is asking the ASX to underwrite a $6 billion AI infrastructure ambition backed by $10 billion of debt.

That is a long way from the "promising startup" stage. Firmus has already pulled in roughly A$1.35 billion of equity in six months, and its valuation jumped from A$1.85 billion post-money in September 2025 to about A$6 billion by November 2025, later confirmed by early-2026 Maas Group capital. Management is now steering toward a roughly A$2 billion IPO during a non-deal roadshow this week. The key question is no longer whether AI infrastructure is real; it is whether private-market valuation momentum can hold up in public markets.

The bull case is straightforward. The market is drawn to a homegrown strategic winner, and the conversation has already stretched to a $6.9 billion valuation framing after Firmus locked in an A$1 billion-plus funding round. Bulls will argue that excitement can keep narrowing the gap between private valuations and the public listing price.

Bears have a simpler objection: sceptics say the optimism is built more on hype than solid foundations. That tension is what defines this listing window.

This is an execution trade, not a pure belief trade in the AI infrastructure narrative.

Why investors are drawn to Firmus: scale, financing, and Nvidia-linked positioning

The bullish case rests on whether Firmus is financing actual build-out, not just branding. The strongest starting point is a US$10 billion debt financing facility led by Blackstone and supported by Coatue. That gives Firmus access to capital at a scale most Australian tech listings never see, which matters in a capital-intensive business where power, sites, and equipment can be as important as early revenue.

The operating case is plausible

Firmus is targeting a modular, liquid-cooled AI factory model built on Nvidia DSX reference architecture, while also positioning itself as an Nvidia Cloud Partner inside Nvidia's Lepton marketplace. Modular design can support faster deployment, liquid cooling fits high-density Nvidia GPU clusters, and Nvidia ecosystem alignment can make the platform easier to sell into existing AI workloads. If that model works, Firmus is not just selling a story about AI demand; it is trying to become infrastructure for it.

Why Australia could matter

Project Southgate is central to that plan. Firmus says it can scale to 1.6 gigawatts of liquid-cooled AI compute by 2028, with a flagship campus planned for 36,000 Nvidia GB300 Grace Blackwell chips and a first stage targeted for 2026. Broader industry commentary around the company also references plans for more than US$70 billion in capital expenditure through 2028. If Australia becomes a credible regional AI hub, the first serious local infrastructure name could capture sovereign demand, export-ready low-carbon compute, and a scarcity premium.

If those pieces connect, this stops looking like financing theatrics and starts looking like a potential infrastructure franchise.

Why the risks still hit hard: float structure, valuation speed, and capex intensity

The main bear case is not that AI demand is fake. It is that public investors could be financing an extremely expensive build-out before the ownership structure makes clear who has exit access.

The float is the clearest warning sign

Reports say the IPO has been pitched with no escrow or share-sale restrictions for early investors. That would be a notable departure from standard IPO practice, where lock-ups help protect new buyers from early holders selling immediately. If early backers can sell without restriction, the listing is not only a ticker launch; it is also a liquidity test.

That concern matters even more because the valuation has moved unusually fast. Firmus went from a A$1.85 billion post-money valuation in September 2025 to roughly A$6 billion by November 2025, with that November level later confirmed by a Maas Group investment in early 2026. Bulls see momentum. Public-market bears see a risk that serial private rounds become the pricing mechanism, leaving newer buyers to fund the bridge to the next stage.

Capex is where the story gets stressed

A US$10 billion debt financing facility is impressive, but it does not remove the fundamental question: how much of that capacity is needed just to fund the build while early equity investors retain implicit exit rights?

Firmus is aiming for 1.6 gigawatts of liquid-cooled AI compute by 2028, while industry commentary around the company references plans for more than US$70 billion in capital expenditure through 2028. That is a huge runway, but it is also a large dilution risk if occupancy, pricing, or delivery disappoint. Sceptics already argue the optimism is built more on hype than solid foundations. When capex needs are this large, that objection matters more.

What would change the assessment

• Positive signal: early investors keep meaningful skin in the game instead of treating the IPO as an immediate exit.
• Negative signal: unrestricted selling lets incumbents turn the listing into exit liquidity.
• What would weaken the bear view: visible commitment from major early holders and evidence that debt access is translating into delivered capacity rather than another financing cycle.
• What would strengthen it: another pre-listing valuation step-up, or fundraising that continues to outrun customer conversion.

What to watch before paying up at IPO

Do not pay the float price first. Wait for proof that this is an asset build, not just a valuation bridge.

The signals that matter at pricing

• Capital deployment matters more than capital availability. The clearest signal is the next phase of Project Southgate moving from financing headlines to physical progress. If construction, permits, power, and customer commitments start to show up, the story is becoming tangible.
• Private-market discipline matters. After three equity rounds in six months and a valuation that quickly trebled to approximately A$6 billion, the healthier outcome would be continued private conviction rather than a forced repricing.
• Ownership structure matters most. If reports of no escrow or share-sale restrictions for early investors are not corrected, treat the IPO as a liquidity event for incumbents rather than a clean handoff to public shareholders.

When the setup improves

• Debt backing is converting into shipped capability, not just bigger balance-sheet optics.
• Early holders keep skin in the game.
• Skeptics shift from hype than solid foundations toward evidence of buildouts and customer traction.

When to stay away