Firmus Withdraws Its A$44 Billion ASX Listing After Investors Wouldn't Meet the Price

Firmus's attempt to become the first Australian AI-infrastructure company on the ASX ended Friday, when the Nvidia-backed data-center builder withdrew its listing application rather than sell shares at a price investors were actually willing to pay. The company says it will now "pursue capital from the private markets and consider alternative public and private market options" -- walking back from a prospectus that had asked public investors to value it at up to A$43.7 billion, roughly eighteen times what the same company was worth thirteen months ago.

The collapse took four days, not four months:

Even the price cut is contested in the reporting -- one account has the revised offer at A$9 a share, most put it at A$8.25 -- but every version of the story agrees on the shape: a syndicate that spent a week trying to find a number the market would actually clear, and failing to find one low enough in the time it had.

Four numbers, four different questions

Part of what made this IPO hard to price is that "Firmus's valuation" has meant something different every time someone's said it this year. None of the four figures below are wrong -- they're just not measuring the same thing, which is exactly the kind of gap a public listing is supposed to close and, this week, didn't:

The gap between the second and third rows is the whole story: a private round two months ago, among investors who already had skin in the game, priced Firmus at $10.5 billion. The IPO asked public investors -- who had never seen the company's books before this prospectus -- to pay more than four times that. eToro's Josh Gilbert put the mismatch plainly: investors wouldn't pay a high price upfront for capacity that's still "largely on the drawing board." Only 46 megawatts of Firmus's planned 911 are actually switched on; the rest is construction, contracts and projection.

The prospectus's own numbers gave skeptics more to work with than usual. Annual revenue today is about A$50 million -- against a valuation ask in the tens of billions. Lock-up restrictions covered only about 42% of shares, meaning early backers could sell large stakes soon after listing, with half the IPO allocation earmarked for those same early holders. University of Sydney professor Michael Biercuk questioned the valuation outright given what he called limited technological innovation in the underlying business. Morgan stockbroker Hugh Robertson was kinder, crediting the founders' "vision and sheer chutzpah" even as he passed on the price.

(Timing may have hurt, too. The float's final week overlapped with a Financial Times report that OpenAI's annualized revenue had come in near $50 billion -- below an earlier $70 billion estimate -- which rattled AI-adjacent stocks generally. One person close to the Firmus deal told The Nightly the link to its own demand was unclear, which is itself an honest answer: a bad week for AI-sector sentiment doesn't need a direct causal chain to matter to a book that was already thin.)

Who actually carries this

A pulled IPO isn't a bankruptcy, and it's worth being specific about who it actually touches:

Having considered recent market volatility and prevailing market conditions, the terms on which the offer could proceed would not appropriately reflect the strength of the company's business, and proceeding with the offer was not in the best interests of the company and its shareholders.

That statement is doing careful work: it blames "market volatility," not Firmus's own numbers, for a price gap that the ledger above suggests was baked in well before this particular week. Firmus has moved fast before -- its private valuation climbed from $1.85 billion at Nvidia's first investment in September 2025 to $10.5 billion in August 2026, each round drawing the same repeat backers rather than new entrants chasing a hot deal. The prospectus itself forecasts US$5.8 billion in EBIT by 2029, built on 865 megawatts still under construction beyond the 46 already operating, and points to roughly A$20 billion in contracted future revenue as the case for today's price. This week answers, for now, the question an earlier round's own "watch" item raised about when a prospectus would finally put those numbers in front of public shareholders: it did, and they didn't buy it at the price asked.

None of this touches the roughly $10 billion debt facility, led by Blackstone's credit arms, that funds Firmus's Indonesia and Malaysia build-out -- a separate instrument from the equity round that just failed. Nor does it change the underlying customer list: OpenAI remains the anchor tenant for two Malaysian facilities, and Meta has commitments in Melbourne and Southeast Asia. What changes is who gets to set the next price. CDC Data Centres' chief strategy officer Jack Dan, speaking generally about the sector rather than Firmus by name, argued some operators chase "financial outcomes" over "business outcomes" and called for government rules around disclosure -- a critique this week's outcome will make harder to wave away.

The company says it will give shareholders more detail "as those options progress," without a timeline. A follow-on private round would let Firmus avoid ever disclosing audited numbers the way a public company must -- the same opacity Jack Dan was pointing at -- while a Nasdaq attempt next year, floated in at least one report, would face the identical revenue-to-valuation gap in front of a different set of skeptical investors. Either path puts off the question this week's buyers just answered: not whether Firmus's build-out is real, but whether its price is.