Nvidia-backed data centre operator Firmus Grid is exploring a private funding round of USD 2 billion to USD 3 billion after abandoning its planned USD 5 billion initial public offering (IPO), as investor concerns over valuations and financing requirements cloud the outlook for artificial intelligence infrastructure companies.
The Australian company and its advisers are in discussions with existing investors about raising the capital, Bloomberg reported, citing people familiar with the matter. The structure of the proposed financing remains undecided, with a combination of equity and debt among the possibilities. It is unclear whether Nvidia will participate in the new round.
Firmus withdrew its application to list on the Australian Securities Exchange on Friday, citing market volatility and prevailing conditions. The company had intended to raise about USD 5 billion at a valuation of approximately USD 30 billion, making the offering one of Australia’s largest share sales in decades.
The withdrawal highlights the difficulties facing AI infrastructure businesses seeking to convert surging demand for computing capacity into valuations that public-market investors are prepared to accept.
Valuation proves difficult to justify
Firmus had marketed its shares at AUSD 11 apiece, targeting a valuation of about AUSD 44 billion, equivalent to roughly USD 31 billion. However, the offer struggled to attract sufficient investor interest, prompting discussions about lowering the price and reducing the size of the offering before the company abandoned the listing.
Investors questioned whether the proposed valuation adequately reflected the company’s operational scale and the substantial investment required to deliver its expansion plans.
John Pearce, chief investment officer at Australian pension fund UniSuper, said the company had an attractive business proposition but that the proposed valuation was difficult to justify.
“So much has to go right to justify the valuation,” Pearce said in an investment update published before the IPO was withdrawn. He also expressed concern that Firmus would need to return repeatedly to capital markets to secure additional debt and equity.
Firmus’s retreat is particularly significant because the company had attracted prominent financial and technology investors. Its backers include Nvidia, trading and technology firm Jane Street, Blackstone and Coatue Management.
In August, Firmus announced a fully committed USD 2 billion strategic equity investment involving existing and new investors, including Nvidia and Blackstone. The transaction valued the business at more than USD 10.5 billion after the investment.
The proposed IPO would therefore have sought a valuation almost three times that level within a relatively short period, intensifying scrutiny of the assumptions underpinning its growth strategy.
AI infrastructure requires enormous capital
Founded in Australia in 2019 as a bitcoin mining business, Firmus subsequently shifted towards data centres and AI computing infrastructure, according to the Financial Times. It aims to develop facilities it calls AI factories, using Nvidia hardware to provide computing capacity for customers developing and running artificial intelligence models.
The company has contracted more than 900 megawatts of capacity but has only 46 megawatts operating across facilities in Australia and Singapore, according to the Financial Times. Five additional data centres are under construction, with the company targeting operations within the next two years.
The gap between contracted capacity and operating infrastructure is central to the investment debate. Delivering the planned expansion will require substantial spending on computing equipment, buildings, electricity connections and cooling systems before all the facilities can generate revenue.
Firmus reported revenue of USD 51 million for its 2026 financial year, according to Bloomberg’s reporting carried by Yahoo Finance. That figure underscores the scale of the challenge in building a business capable of supporting a multibillion-dollar valuation.
The company operates in the fast-growing neocloud segment, in which specialist providers supply computing power to AI developers rather than competing across the full range of cloud services offered by Amazon, Microsoft and Google.
Demand for AI computing has created opportunities for these providers, but their expansion models can depend heavily on borrowed money and continued access to equity markets.
Debt and investor caution
Analysts have warned that the financing requirements of AI infrastructure businesses could become a vulnerability if revenue growth falls short of expectations or borrowing costs remain elevated.
Morningstar senior market strategist Lochlan Holloway has highlighted the leverage involved in the neocloud model, under which companies borrow against customer contracts to purchase chips and use rental income to service their loans.
Firmus expects to carry about USD 30 billion in debt once its planned data centres are built, equivalent to roughly six times the USD 5 billion in operating earnings it forecasts for 2028, according to ABC News. These are projections rather than current financial results, but they illustrate the scale of the future financing commitment.
The IPO’s failure also comes amid wider questions about whether the enthusiasm surrounding artificial intelligence has pushed infrastructure valuations beyond what operating businesses can presently support.
Firmus has said it will pursue private-market funding while considering other public and private options. A successful private round could provide capital to continue construction while avoiding the valuation investors rejected in the public market. However, the terms of any deal will determine how much funding the company secures and at what valuation.
The immediate priority is to finance its expansion while demonstrating that contracted demand can translate into operating capacity and sustainable earnings.
For the wider AI industry, Firmus’s withdrawal is a reminder that strong demand for computing power does not automatically guarantee investor appetite at any price. The next phase of the AI infrastructure boom may depend as much on disciplined financing and credible execution as on technological ambition.