The spectacular failure of Firmus Technologies should not have come as a surprise. It was a float of hope as opposed to reality. With a valuation wildly above earnings the market was left asking the question is how can a company earning $70 million really be worth $44 billion?
Australia’s hopes of hosting one of the largest initial public offerings in its history were abruptly derailed on 9 October when Firmus Technologies withdrew its application to list on the Australian Securities Exchange (ASX). The company had been preparing to raise billions of dollars to finance its expansion into artificial intelligence data centres, with an anticipated valuation approaching $44 billion.
The collapse of the proposed float followed disappointing investor demand, mounting questions about the company’s valuation and concerns over how quickly its ambitious infrastructure plans could be delivered. What had been promoted as a landmark opportunity to invest in the AI revolution instead became a cautionary tale about the gap between technological potential and commercial reality.
A giant valuation built on future growth
Firmus was founded in 2019 by Oliver Curtis, Tim Rosenfield and Jonathan Levee. Originally involved in cryptocurrency mining, the business subsequently shifted towards providing computing infrastructure for artificial intelligence.
Its strategy centres on building specialised data centres capable of accommodating the powerful processors required to train and operate AI systems. The company has attracted high-profile investors, including American chipmaker Nvidia, alongside global investment firms and other institutional backers.
The attraction is understandable. The rapid expansion of generative AI has created enormous demand for computing power, electricity, cooling systems and data storage. Technology companies are competing to secure infrastructure capable of supporting increasingly sophisticated AI models.
Firmus sought to capitalise on that demand by proposing a multibillion-dollar sharemarket float. Yet its intended valuation of approximately $44 billion raised a fundamental question: how much should investors pay today for infrastructure and earnings that may take years to materialise?
At that valuation, Firmus would have ranked among Australia’s largest listed companies, despite having a comparatively limited operational footprint relative to its planned expansion.
The problem was not necessarily that AI infrastructure lacked commercial potential. It was whether the company’s prospective earnings could justify the price being demanded from investors.
Investors fob off exorbitant asking price
The offering unravelled during the institutional bookbuilding process, when investment banks gauge demand and determine the price at which shares can be sold.
According to reporting by The Guardian, the proposed offer price initially stood at $11 a share before repeated attempts to reprice the deal. The proposed price was reportedly cut substantially as the company and its advisers tried to attract sufficient demand. Even those efforts failed to rescue the transaction.
The withdrawal demonstrated the limits of investor enthusiasm for AI-related businesses. While investors may be prepared to pay premiums for companies exposed to a rapidly expanding industry, they still need confidence in the underlying financial projections, funding requirements and execution risks.
Firmus attributed its decision to recent market volatility and prevailing market conditions.
Its board concluded that the terms on which the offering could proceed would not adequately reflect the company’s business and long-term growth outlook.
The company subsequently indicated it would pursue private-market funding and consider alternative public and private financing options. A future overseas listing has also been discussed in reporting about its plans.
The infrastructure delivery problem
One of the central concerns confronting prospective investors was the difference between Firmus’s planned capacity and infrastructure already operating.
The Financial Times reported that Firmus had contracted more than 900 megawatts of capacity but had only approximately 46 megawatts operating across facilities in Australia and Singapore.
That distinction is crucial. Contracts, development proposals and future capacity targets can represent substantial commercial opportunities, but they are not equivalent to completed facilities generating established revenue.
Building AI data centres requires enormous capital expenditure, access to electricity, reliable construction partners, specialised cooling systems and expensive computing equipment. Delays can increase costs and postpone revenue, while higher interest rates make debt-financed expansion more expensive.
There were also questions about the future of Project Southgate, an ambitious Australian data centre development associated with infrastructure provider CDC Data Centres. Industry reporting indicated that the partnership had ended after only a small portion of the proposed capacity was delivered. Firmus maintained that the decision did not undermine its wider development plans.
These developments intensified scrutiny of the company’s ability to convert its long-term ambitions into operating infrastructure.
An uncomfortable question about AI valuations
Firmus’s failed float also reflects a broader debate about whether the AI investment boom has moved ahead of commercial fundamentals.
The world’s largest technology companies are committing unprecedented sums to computing infrastructure. However, the financial returns from these investments remain uncertain, particularly when facilities must be built before customers can use them and before the full costs of operating them become apparent.
Investors must consider whether demand will continue growing fast enough to justify current construction plans, whether computing capacity will become cheaper and whether technological advances could change the economics of existing facilities.
Competition is another risk. Data centre operators must compete for customers, electricity supplies, suitable sites and financing. Even where demand is strong, not every company will necessarily earn attractive returns.
The Firmus experience suggests that institutional investors are increasingly distinguishing between companies that already generate substantial cash flow and those whose valuations depend heavily on future execution.
Lessons to be learned
The failed IPO represents a setback for Australia’s ambitions to attract major technology listings and deepen its capital markets. A successful offering could have demonstrated that the ASX was capable of financing a globally significant AI infrastructure business.
Instead, the withdrawal raises questions about how much risk public investors are willing to accept when funding capital-intensive technology ventures.
For Firmus, the immediate challenge is to secure the money needed to continue its expansion without the proceeds of the proposed float. Private investors may still be willing to provide capital, but they could demand different valuations, stronger contractual protections or clearer evidence of progress.
For investors, the lesson is more fundamental. An exciting industry, prominent financial backers and impressive growth forecasts do not eliminate the need to examine a company’s balance sheet, cash flow, construction commitments and realistic path to profitability.
Firmus’s IPO failure does not prove that AI infrastructure is a bad investment. It demonstrates that even in a sector attracting enormous global investment, valuation and execution remain decisive.
The AI revolution may ultimately create highly profitable businesses. But Firmus has shown that the public market will not necessarily finance every ambitious plan at any price.
