- Firmus Grid pulled what would have been one of Australia’s largest IPOs after US fund managers refused the $30 billion valuation. Bankers floated cutting the deal to about $3 billion at a $20 billion to $25 billion valuation, and investors declined that too.
- The arithmetic explains the refusal. Firmus generated $51 million of revenue in the 2026 financial year, so the $30 billion ask represented roughly 588 times trailing revenue, and of its 912 megawatt pipeline just 46 megawatts had been built, around 5%.
- The valuation had moved extraordinarily fast. Portfolio manager Leonid Mironov of Gavekal Capital noted it was worth under $2 billion a year ago, $10 billion in August after a $2 billion round backed by Jane Street and Blackstone, and was asking three times that in October. An April round with Coatue and Nvidia valued it at $5.5 billion.
- Contagion was immediate for a listed backer, with Maas Group Holdings falling as much as 30% in Sydney, its largest decline on record. Firmus is now exploring a smaller private round with existing investors, who were set to buy half the IPO shares in any case.
What Happened?
Firmus pitched an enterprise value to EBIT multiple of 13 times, framed as a significant discount to US peer CoreWeave, but based on earnings forecasts two years ahead, while CoreWeave has both a longer record and revenue orders of magnitude larger. Concerns had been building publicly after it emerged existing shareholders would not be subject to escrow and could sell freely. UniSuper, one of Australia’s largest pension funds, said as early as July it would not participate because too little was known about the business. Days before pricing, Firmus announced an agreement with existing customer Meta for computing capacity in Southeast Asia and revised its numbers upward on that basis, which some investors saw as pushing the envelope. Co-founder Oliver Curtis served a prison sentence for insider trading a decade ago, which Bloomberg reports was not cited as a deal-breaker but added caution. Jun Bei Liu of Ten Cap Investment said she had never seen an IPO so polarising.
Why It Matters?
The forward multiple is the mechanism by which an extreme valuation was made to sound reasonable, and the lesson generalises well beyond this deal. Thirteen times EBIT sounds disciplined, particularly framed as a discount to a listed peer, until you note the earnings are forecast two years out for a company that has built 5% of its pipeline. Any valuation becomes a modest multiple if you project far enough forward. The trailing figure, 588 times revenue, is the one that survives contact with scrutiny, and whenever a forward multiple is quoted at a discount to a comparable, that is the number to ask for. This is also the first genuine refusal in a market that has absorbed everything else. SpaceX is seeking $40 billion for chips, an AI cloud provider is placing a $5 billion loan at 11%, a single bank has written roughly $35 billion of acquisition bridges in days, and a loss-making data centre developer rose on quadrupled revenue. Against that, institutional investors declined this deal at $30 billion and declined again at $20 billion to $25 billion. A boundary exists, and it has now been located. The question is whether it applies only to companies without a track record or marks a broader change in appetite. The process failure deserves attention from anyone participating in bookbuilds. Firmus claimed strong strategic and global demand while bankers reported indications exceeding the offer size, and the deal collapsed within 48 hours, which Bloomberg notes raises questions about how the company and its advisers read demand so badly. Adding a customer agreement days before pricing to justify a higher number is precisely the behaviour that erodes confidence in those signals. On marks, be careful about the read-across: the August round priced at $10.5 billion and the market refused $20 billion to $25 billion, so those investors are not necessarily underwater, they simply will not receive the markup the IPO would have delivered.
What Next?
The private round with existing investors is the immediate development, and the valuation it clears at will show what Firmus is actually worth to people who already own it. Watch whether other AI infrastructure listings are delayed or repriced, since this is the clearest signal yet that public investors are distinguishing between companies with revenue and companies with pipelines. Maas Group’s recovery or otherwise is the visible contagion channel. For the sector, CoreWeave’s multiple now carries more weight as a reference point given it was the benchmark used here. The broader question is whether private AI infrastructure marks, which have risen very quickly across the sector, face the same scrutiny when those companies eventually seek public capital.
Affected Tickers and Coins: NVDA, MGH, BX, META, CRWV
Source: Bloomberg
