NVIDIA (NVDA.US)-backed Australian artificial intelligence (AI) data center operator Firmus Grid had originally planned to list at a valuation of about $30 billion, which would have made it one of the largest initial public offerings (IPOs) in Australia. However, due to an excessively high valuation, insufficient investor subscription demand, and market concerns over the massive financing needs of AI infrastructure, the closely watched listing plan was ultimately canceled on Friday.

According to people familiar with the matter, Firmus and its underwriting team had hoped to attract a large number of U.S. institutional investors to participate in the subscription, but during the roadshow from Tuesday to Wednesday of this week, several U.S. fund managers considered the $30 billion valuation the company was seeking to be too expensive. As subscription demand fell short of expectations, the underwriting team at one point considered reducing the size of the offering and lowering the valuation in an effort to salvage the deal. It is understood that the relevant plans included cutting the fundraising scale to about $3 billion and lowering the company's valuation to between $20 billion and $25 billion. However, even after adjusting the terms, the underwriting team still failed to secure enough support from U.S. investors and ultimately had to cancel the IPO.

Bank of America (BAC.US), JPMorgan Chase (JPM.US), Morgan Stanley (MS.US), and Australian financial services firm Morgans Financial served as joint lead underwriters for the IPO. In fact, Firmus's listing plan had previously sent positive signals. When finalizing the offering plan, the company claimed it had received strong demand from strategic investors and global institutional investors, and the underwriters also said preliminary subscription indications exceeded the offering size. But as the listing process advanced, questions about the company's valuation and deal structure gradually intensified.

One important issue was that existing shareholders might not be subject to share lock-up agreements, meaning they would have an opportunity to quickly sell their holdings after the company went public, potentially putting pressure on the stock price. On Wednesday of this week, news of weak subscription demand and a possible cut to the offering price emerged one after another, further intensifying investor unease. Some fund managers began reducing their subscription orders, while others directly withdrew their subscription intentions. On Thursday morning, Firmus's IPO subscription process ended as scheduled, but the market remained uncertain whether the company could complete the offering at the original price of A$11 per share.

Affected by the related concerns, shares of Maas Group Holdings, one of Firmus's supporters, plunged as much as 30% in the Sydney market at one point, the largest intraday drop on record. Jun Bei Liu, co-founder and chief portfolio manager of Australian investment institution Ten Cap Investment, said she had never seen an IPO that triggered such severe market disagreement. She also noted that if the listing failed, Firmus might instead seek financing from existing investors.

The sharp rise in valuation over a short period became one of the core reasons why Firmus's listing was blocked. In April of this year, Firmus was valued at about $5.5 billion in a funding round backed by Coatue Management and NVIDIA. By August, the company completed another $2 billion funding round, attracting well-known investment institutions such as Jane Street and Blackstone (BX.US), and its valuation rose to more than $10.5 billion. However, just two months later, Firmus sought to enter the public market at a valuation of about $30 billion, nearly tripling its valuation from the August funding round. Such a rapid increase in valuation raised questions among some investors about whether the company's actual operating scale could support its market value.

According to disclosed data, Firmus's revenue for fiscal year 2026 was only $51 million, while the valuation sought in this IPO exceeded $30 billion. At the same time, the company's planned total data center capacity reached 912 megawatts, but only 46 megawatts had actually been built. This means investors would need to value the company based on a large number of data center projects not yet built and on profits that might be realized in future years. Whether these growth targets can be achieved still depends on whether project construction, customer demand, and massive financing can be smoothly carried out.

During the IPO roadshow, Firmus used the ratio of enterprise value to earnings before interest and taxes (EV/EBIT) as its valuation metric and compared itself with companies such as U.S. AI cloud computing company CoreWeave (CRWV.US). Although the roughly 13-times valuation multiple proposed by Firmus was significantly lower than that of CoreWeave, this calculation was based on expected profits two years later rather than the company's current actual earnings performance. By comparison, CoreWeave has a longer operating history, and its revenue scale is far higher than Firmus's.

Leonid Mironov, portfolio manager at investment institution Gavekal Capital, said bluntly that the IPO had multiple problems, including a poor deal structure, an excessively large offering size, overly high pricing, and unreasonable valuation metrics. He pointed out that Firmus's valuation was less than $2 billion less than a year ago, rose to about $10 billion in August, and by October was trying to list at a valuation three times that level. The report said some investors could originally have accepted a valuation of about $25 billion, but just days before IPO pricing, Firmus announced a computing capacity supply agreement with existing customer Meta Platforms (META.US), involving its data center facilities in Southeast Asia. The company then raised its relevant financial forecasts based on this agreement and set a higher IPO valuation accordingly.

This adjustment further deepened the doubts of some investors. They believed that Firmus was still in the early stage of business development but was asking investors to pay a high premium in advance for high growth over the coming years, while achieving those growth targets would also require billions of dollars in funding. In addition, the personal history of Firmus co-founder Oliver Curtis also drew attention. Curtis had served a prison sentence about a decade ago for insider trading. Although people familiar with the matter said this was not a decisive factor in the IPO's failure, the background still added to caution among some investors.

The sudden failure of Firmus's listing plan also reflected rising investor attention to the financing risks of AI infrastructure. As global technology companies accelerate data center construction, related projects require enormous amounts of capital, but it often takes a long time from completion of construction to generating stable revenue. For companies that lack a mature operating record and still rely on future projects to deliver growth targets, investors are beginning to scrutinize their valuations, funding needs, and execution capabilities more strictly.

Ultimately, because subscription demand from U.S. institutional investors fell short of expectations and demand from Australian domestic investors was not enough to fill the gap, Firmus formally withdrew its IPO plan on Friday. According to people familiar with the matter, the company is now considering turning to smaller private financing and may raise about $3 billion from existing investors. Previously, these existing investors had already been prepared to subscribe for about half of the shares offered in the IPO.