Firmus's $5B IPO Faces the 'Fairytale' Valuation Question

By Daniel Rao

Photo: Melbourne, where Australian AI data centre operator Firmus runs existing facilities. The company has priced a A$7.1 billion IPO — set to be the country's second-largest listing ever — at a valuation that has nearly tripled in under two months. (Bob Tan, via Wikimedia Commons, CC BY 4.0)

Australian data centre operator Firmus has priced what would be the country's second-largest IPO in history. Its valuation has nearly tripled in less than two months. And the fund managers being asked to buy it are openly calling the company "unproven."

That combination — a record-breaking listing and a buyer strike at the same time — is what makes the Firmus float the most interesting financial story in AI infrastructure this month. Thursday's pricing, per a term sheet seen by Reuters, went in at A$11 a share — roughly A$7.1 billion (about US$5 billion) for an equity valuation of $30.6 billion. Back at the start of August, a funding round had priced the same company at $10.5 billion. The market has re-rated Firmus by nearly 200% in under eight weeks. The people who manage money for a living are looking at that number and asking, with unusual candour, what exactly has changed.

Quite a lot, it turns out, is being asked to change. Nothing else, yet.

The A$44 billion number that made fund managers blink

Start with the raw terms, because they are extraordinary by any standard. A$11 a share ($7.63) puts Firmus on course for the biggest Australian listing in recent memory — second only to one earlier deal in the country's history, per Reuters. Bookbuilding kicks off Tuesday; the ASX debut is pencilled in for 23 October. The term sheet says indicative orders are already above the deal's size — so for all the scepticism you're about to read below, the FOMO is very much real.

The valuation arc is the story. $10.5 billion at the start of August; $30.6 billion now. We ran the arithmetic: that is a roughly 190% re-rating in under eight weeks, during which the company built no new data centres, opened no new revenue lines and, as far as anyone can see, did not discover a new continent of computing demand. What it did was attach itself, ever more tightly, to the AI infrastructure narrative at precisely the moment that narrative is absorbing every spare dollar in global capital markets.

And the leverage. Merlon Capital Partners portfolio manager Kirit Hara, one of the sceptics who spoke to Reuters, flagged that the company's debt pile sits at around $30 billion — a figure Morningstar strategist Lochlan Halloway put at roughly six times forecast earnings. Six times. For a data centre operator. In a rising-rate environment.

Who's in the tent

Firmus is not some garage startup with a dream and a pitch deck. Its backers read like a roll call of the AI buildout's winners: Blackstone, Nvidia, and local institutions including Wilson Asset Management and Argo. An earlier strategic equity investment of US$2 billion involved Coatue, Nvidia, Blackstone and Jane Street, aimed at expanding Firmus's AI Factory infrastructure across Australia and the wider Asia-Pacific region.

That Nvidia connection is doing enormous work in this story. Nvidia's imprimatur has become the closest thing the AI economy has to a credit rating — when the chipmaker blesses an infrastructure play, the market reads it as a demand signal from the one company that sees every order book in the industry. Jensen Huang's company does not need Firmus to succeed; but Firmus very much needs the market to believe Nvidia needs Firmus.

Photo: Nvidia CEO Jensen Huang at Computex Taipei. Nvidia is among Firmus's backers, and its imprimatur is doing heavy lifting in the IPO's valuation story. (NVIDIA Taiwan, via Wikimedia Commons, CC BY 2.0)

The company's promise

Firmus's pitch is scale, and the numbers are genuinely ambitious. The company estimates that five planned data centres across Asia-Pacific, together with its existing facilities in Melbourne and Singapore, will generate some $5 billion in annual earnings within five years, Reuters reported last month. Read that again: $5 billion of annual earnings, within five years, from a company whose current operating footprint is two sites.

We ran that multiple too. $30.6 billion of equity value against $5 billion of earnings — five years from now, if everything goes right — is a ~6x multiple on hope, before the $30 billion debt pile enters the room. Infrastructure investors normally pay single-digit multiples for assets that already exist. Here they are being asked to pay a growth multiple for assets that mostly do not.

It is the same financing creativity reshaping the whole AI stack. Amazon is exploring a structure to move $8 billion of Nvidia chips off its books into an investor-funded vehicle, because even the richest companies on earth are straining to carry this buildout on their own balance sheets. Firmus is the purest expression of the trend: an asset-heavy, debt-fuelled bet that AI compute demand will grow into any valuation, no matter how stretched.

And the scramble is not just financial — Tencent is leasing 100,000 chips it can't legally buy, because the whole industry is fighting over the same finite pool of compute.

There is a physical constraint, too — Britain's grid is already groaning under data-centre demand, and every new campus needs power as well as capital.

The arithmetic that doesn't add up

Now the quotes, because they are unusually blunt for fund managers talking about a live deal. Hara, of Merlon Capital Partners, told Reuters: "Our process prevents us from effectively buying into, kind of, the hopes and dreams." Then the kicker: "We're anchored towards what's actually on the table. And what's on the table in terms of operating and development, we're struggling to get to that kind of A$40-odd billion dollar market cap." He added that the company's $30 billion of debt was a concern given rising costs and interest rates.

"Hopes and dreams" is not language you use about a deal you expect to participate in. It is the language of a buyer who has done the work and walked away.

Halloway, the Morningstar strategist, said in a note that Firmus had the hallmarks of a "boom phase," pointing to the "wild increase in valuation in such a short period of time." "Boom phase" is Morningstar-speak for "this is what the top looks like," and Halloway did not leave it ambiguous.

Romano Sala Tenna, a portfolio manager at Katana Asset Management, told Reuters the IPO was polarising potential investors and that he was still undecided: "We are struggling with the fundamental arithmetic, and that there is an enormous amount of execution which is required to justify the valuation." Katana's criteria — cash flow, earnings — are not met, he said, though the stock could work as a trade if passive fund flows create demand after listing. Translation: the fundamentals don't justify it, but the index bid might.

And Joseph Koh of Sydney's Blackwattle Investment Partners was the bluntest of all: his firm would not be bidding. "Because we don't have enough confidence in the delivery of the future projects, we can't get any confidence on the valuation that underpins that," he told Reuters. Blackwattle prefers its AI exposure through Goodman Group, which Koh said has a better track record of developing data centres on time and on budget than "unproven" Firmus. "There are so many unknowns," he added — before the obligatory hedge: "But that doesn't mean we would look to short it either."

Four named sceptics, three firms passing, one "boom phase" call. And indicative orders already above the deal size. Both things are true, and the tension between them is the entire story of the AI trade in 2026.

The bull case — and it isn't nothing

Fairness requires the other side. The bull case for Firmus is not that the valuation is reasonable; it is that reasonableness stopped mattering in AI infrastructure some time ago. Demand for AI compute is real and violently growing. Hyperscalers are signing decade-long capacity contracts. Sovereign governments — Australia's included — are treating domestic AI infrastructure as a strategic asset. A company with land, power, Nvidia's blessing and Blackstone's balance-sheet creativity is holding genuinely scarce cards.

The oversubscribed book is the market's answer to the sceptics: somebody always pays up. Passive flows, FOMO, the Nvidia halo, the sheer scarcity of listed AI-infrastructure exposure in Australia — any one of these can carry a deal that the fundamentals can't. Michael Burry has swapped his AI shorts for puts, betting the bubble bursts on a timetable; the Firmus bookrunners are betting it doesn't burst before 23 October. Both can be rational. Only one gets to be right.

There is also a real business underneath the multiple. Two operating facilities, five planned campuses, an existing footprint in two of Asia-Pacific's most important markets, and a management team that has already raised $2 billion of strategic equity from the smartest money in the sector. "Unproven" is fair. "Nothing" is not.

Firmus IPO: the questions everyone asks

What is Firmus worth in the IPO?
Firmus has priced its IPO at A$11 a share ($7.63), which implies an equity valuation of roughly A$44 billion — about US$30.6 billion. That is 44 times the roughly $1B valuation the company carried 12 months ago, according to the Oct 2 term sheet reported by Reuters.

When does Firmus list on the ASX?
The institutional bookbuild runs Oct 6–9, with trading expected to begin on the ASX on Oct 23. Indicative orders already exceed the size of the offer, per the term sheet.

Is Firmus profitable?
The IPO pitch is about future earnings, not current profits: the company estimates its five planned Asia-Pacific data centres, plus existing facilities in Melbourne and Singapore, will generate $5 billion in annual earnings within five years. That is precisely what the sceptics dispute — Morningstar's Lochlan Halloway put the company's $30 billion debt pile at roughly six times forecast earnings.

Who backs Firmus?
Blackstone, Nvidia, Coatue and Jane Street took part in a US$2 billion strategic equity investment, and local institutions including Wilson Asset Management and Argo are in the tent. Firmus has not publicly named its anchor customers in the term-sheet reporting.

What are the biggest risks?
Execution (five planned campuses — "an enormous amount of execution which is required," in Katana's Romano Sala Tenna's words), leverage ($30 billion of debt in a rising-rate environment), and valuation — Marcus Today put fair value at $32–45 billion with a High risk rating. There is also local backlash: Tasmanian resident Kayla Thompson told Guardian Australia neighbours were unaware of the local project until construction started.

Update, Oct. 3

The Oct 2 term sheet, reported by Reuters, firms up the numbers behind this story: A$11 a share, an equity valuation of about A$44 billion — 44 times where Firmus stood 12 months ago — with the institutional bookbuild running Oct 6–9 and ASX trading from Oct 23. Indicative orders already exceed the offer. The scepticism has not softened: one fund manager called the forecasts "a little bit of a fairytale," while another told Guardian Australia "these guys are just coming from nothing." In Tasmania, resident Kayla Thompson said neighbours were unaware of the local project until construction started, and that her stepchildren can see the site from their classroom window. Marcus Today put fair value at $32–45 billion with a High risk rating, noting Firmus's bitcoin-mining origins. Firmus declined comment to Reuters.

Our take: this is the AI boom's honesty test

Here is my position, stated plainly. The Firmus IPO is the most honest document the AI boom has produced this year — not because the company is honest (its projections may prove exactly right), but because the deal forces everyone to say the quiet part out loud. The sceptics are saying it: "hopes and dreams," "boom phase," "struggling with the fundamental arithmetic." The buyers are saying it too, with their wallets: indicative orders above the deal size, for a company whose debt is six times forecast earnings. Both sides are describing the same trade. One side calls it infrastructure; the other calls it a greater-fool relay. The uncomfortable reality is that AI infrastructure has been both, simultaneously, for two years now — OpenAI's own pre-IPO ambitions run to the trillions — and the only thing that decides which description wins is whether the demand shows up on schedule.

What worries me is not the valuation. Valuations are opinions with marketing budgets. What worries me is the leverage: $30 billion of debt, six times forecast earnings, in a sector where the "forecast" part of "forecast earnings" is doing all the work. Debt does not care about narratives. Debt has covenants, maturities and interest bills, and it collects them whether or not the five new data centres open on time. If Firmus executes, this IPO will look like the bargain of the decade. If it stumbles — a delayed campus, a softening in compute demand, a rate shock — the equity cushion underneath that debt starts to look very thin, very fast.

The history of infrastructure booms is not subtle about what happens next. The assets get built, the demand mostly arrives, and the equity returns concentrate in the hands of whoever bought after the first restructuring. The question for anyone considering this IPO is not whether AI needs data centres. It does. The question is whether it needs them at $30.6 billion, from this company, right now — and whether you're being paid for the risk that the answer is "eventually."

What we're watching next

Three things. First, the bookbuild starting Tuesday: does the scepticism in the press translate into a soft book, or do the indicative orders hold and the deal prices at the top? Second, the 23 October listing — first-day trading will be the market's real verdict, and "boom phase" IPOs have a habit of looking invincible until they don't. Third, the debt: $30 billion at six times forecast earnings leaves no room for a stumble, so every construction update on those five planned campuses becomes a credit event in miniature.

The honest headline out of Sydney this week is not that Australia is getting its biggest-ever tech listing. It is that the AI boom has reached the stage where the sceptics and the buyers are the same story — and both of them are telling the truth.

Sources

Reuters reporting on the Firmus IPO terms, valuation, backers and investor scepticism — including the A$11 ($7.63) pricing, $30.6 billion equity valuation, $10.5 billion August valuation, second-largest Australian IPO, five planned Asia-Pacific data centres plus existing Melbourne and Singapore facilities, $5 billion five-year earnings target, $30 billion debt pile, and named quotes from Kirit Hara (Merlon Capital Partners), Lochlan Halloway (Morningstar), Romano Sala Tenna (Katana Asset Management) and Joseph Koh (Blackwattle Investment Partners) — 2 October 2026; Informax Prime on the US$2 billion strategic equity investment involving Coatue, Nvidia, Blackstone and Jane Street, the A$7.1 billion raise target, October bookbuilding and 23 October ASX trading date. The ~190% re-rating and ~6x forward-multiple calculations are TechPulse USA's own arithmetic from the reported figures. $1 = 1.4413 Australian dollars, per Reuters.

Daniel Rao covers AI safety, frontier labs and technology policy for TechPulse USA. He writes about the people and institutions deciding what artificial intelligence is allowed to do.