Australian data center operator Firmus is preparing to launch what could become one of the largest initial public offerings in the country’s history, seeking to raise A$7 billion ($5 billion) as investors pour capital into the infrastructure needed to support the global artificial intelligence boom.
According to a term sheet reviewed by Reuters, Firmus plans to begin its institutional bookbuild on October 6 and close it on October 7, with the company’s prospectus expected to be lodged on October 8. Retail investors will be able to bid for shares from October 12 to October 19, while trading on the Australian Securities Exchange is scheduled to begin on October 22.
The offering would rank as the second-largest IPO in Australian history, behind Telstra’s A$10 billion listing in 1997. It would also rank as the world’s fourth-largest IPO so far this year, according to Dealogic data.
The scale of the proposed listing makes Firmus an important test of how investors are valuing companies that provide the physical infrastructure required for AI computing. The sector has attracted enormous investment as technology companies race to secure data centre capacity, power and advanced computing equipment, but concerns are growing over whether the pace of spending can be sustained.
Firmus, founded in 2019 by Oliver Curtis, Tim Rosenfield and Jonathan Levee, could raise as much as A$7.7 billion if an over-allotment option, or greenshoe, is exercised. The option would allow the company to raise an additional A$500 million.
The founders will remain subject to escrow arrangements after the listing, according to an investor presentation reviewed by Reuters. Only 10% of their shares would be released after one year, followed by a further 39.9% after two years.
The lock-up structure limits the amount of founder-held stock that can immediately enter the market, potentially providing greater visibility over the company’s shareholder base during the early stages of its life as a public company.
Firmus currently operates two data centers in Australia and Singapore and has another five facilities under development across the Asia-Pacific region. The expansion comes as demand for computing infrastructure accelerates alongside the rapid deployment of AI models and applications.
Data centers have become one of the biggest beneficiaries of the AI investment cycle. The growth of large language models and AI agents requires substantial computing capacity, creating demand not only for chips but also for facilities capable of housing and powering them.
That demand has helped turn data center capacity into a scarce infrastructure asset. Operators are competing for access to electricity, land, and network connections, while technology companies are increasingly signing long-term agreements to secure capacity before new facilities are completed.
But the same expansion that has created an opportunity for operators such as Firmus is generating a growing set of constraints.
Data centers consume large amounts of electricity and, depending on their cooling systems, significant quantities of water. Communities in Australia and other markets have been pushing back against new developments over concerns about their impact on local power grids, water supplies and surrounding areas.
Firmus itself has faced local opposition to planned projects in Australia, highlighting a central problem for the sector: securing financing may be easier than securing the physical resources and community support needed to turn that financing into operational capacity.
The company is also entering public markets at a time when some technology executives and investors are questioning the durability of the AI infrastructure boom. Billions of dollars are being committed to new facilities on the assumption that demand for AI computing will continue to grow rapidly, but the economics ultimately depend on technology companies continuing to spend heavily on training and running capable models.
The situation has resulted in a different risk profile from conventional infrastructure. Data centers can have long operating lives, but the pace of AI development means the type, location, and scale of computing demand can change rapidly.
Therefore, Firmus’ IPO is expected to provide a market test of whether the infrastructure layer of the AI economy can command valuations comparable to the technology companies driving demand for it.
The proposed offering also shows how the AI investment cycle is broadening beyond semiconductor manufacturers and software developers. Capital is increasingly flowing into the physical systems required to support AI, including data centers, power generation, cooling infrastructure and specialized computing facilities.
Firmus’ Asia-Pacific expansion gives it exposure to that broader trend, while also placing the company in markets where access to electricity and suitable land can become constraints on development.
The size of the IPO means its performance after listing could become an important signal for other companies seeking to raise capital for AI-related infrastructure. Industry analysts believe a strong reception would demonstrate that public-market investors remain willing to finance large data center expansion programmes even as concerns about AI spending grow. A weaker reception could make it harder for other infrastructure operators to raise money at the valuations they have been targeting.
However, the immediate challenge for Firmus is to convince investors that its growth plans are supported by durable demand rather than simply by the current surge in AI spending.
The company’s planned October 22 listing will put that question directly to the market. Its success will depend not only on how quickly AI demand grows, but on whether Firmus can convert that demand into long-term contracts, operating capacity and returns while navigating the difficult competition for power, land and public acceptance.
