Australian data-centre operator Firmus has shelved its planned US$5 billion initial public offering, a decision that underscores how even heavyweight backing from Nvidia cannot persuade public investors to accept any price for an artificial intelligence story. The company abandoned the listing on Friday after demand weakened, ending what would have been the country’s second-largest float.
Firmus had marketed shares at A$11 each, implying an equity valuation of roughly US$30.6 billion—nearly three times the US$10.5 billion valuation from its August funding round. The company attributed the withdrawal to market volatility and said it would instead pursue private capital. But investor commentary suggests the real issue was that the price already reflected too much future growth before most of the infrastructure needed to deliver it had been built.
Nvidia’s credibility could not overcome the price
Firmus arrived with an impressive roster of backers, including Nvidia, Blackstone, Coatue Management, and Jane Street. The AI boom has fueled demand for data centres and power, giving the company a compelling narrative. Yet that was not enough to justify the valuation.
UniSuper chief investment officer John Pearce noted in an investment update that Firmus had “a compelling story” but “doesn’t have a compelling valuation.” He added that the company was “priced to perfection,” requiring too many things to go right for the proposed valuation to make sense. The concern was amplified by the speed of the valuation jump—from US$10.5 billion to US$30.6 billion in less than two months.
Joseph Koh, portfolio manager at Blackwattle Investment Partners, who reviewed the deal but did not bid, told Reuters that investors were being asked to pay a large price for future outcomes assuming “near flawless execution.”
Most revenue depends on facilities yet to be built
Firmus currently operates two leased data centres, in Melbourne and Singapore, and plans five more across the Asia-Pacific. Its draft prospectus projected about US$5 billion in annual earnings within five years—an ambitious forecast given that much of the capacity underpinning it remains under development.
Minotaur Capital co-founder Armina Rosenberg told The Guardian that about 97% of contracted revenue was linked to sites that had not yet been built. “You only get near the offer price if delivery, financing and renewals all go to plan,” she said. That leaves investors underwriting multiple risks simultaneously: construction timetables, access to power, financing costs, customer demand, and contract renewals. Any slippage could materially alter the earnings profile.
Rising borrowing costs shift the AI financing debate
Long-term bond yields have climbed sharply, increasing the cost of financing capital-intensive projects as investors become more demanding about returns from AI spending. Firmus is expected to require additional debt and equity as it expands—a point Pearce highlighted when explaining UniSuper’s reluctance to participate.
The failed float also comes as markets question the scale of funding being sought across the AI ecosystem. Huge infrastructure programmes can create enormous value, but investors increasingly want evidence that spending will convert into durable cash flow. The Firmus episode serves as a cautionary tale for other AI-related companies considering public listings at lofty valuations.
This article is for informational purposes only and does not constitute financial advice.
