Firmus Cancels $11-a-Share IPO as Investors Balk at AI Data Centers

Firmus Technologies has abandoned its share offering on the Australian Securities Exchange (ASX). The offer price was $11 per share, and the deal would have been the largest market debut since Telstra in 1997. The reason is stated plainly: investor demand for the AI data center business, which has attracted so much talk, never came together. A company representative said the board considered continuing the deal contrary to the interests of the company and its shareholders.

What we know about the deal

In IPO practice, a "best interests" statement usually means the order book either did not reach the desired size or reached it only at a price the issuer and the underwriting banks were unwilling to accept. The published announcement gives no final order volume, no size of the planned raise, and no names of anchor investors. The exact scale of the failure therefore cannot be judged. What is known is the key point: the deal was not merely delayed but called off.

The comparison with Telstra is no accident. The 1997 privatization of the telecom giant was a mass offering that drew hundreds of thousands of retail investors. That is a high bar. Firmus was aiming for a comparable symbolic status, but the market was not ready for that role.

• Offer price: $11 per share.
• Deal status: cancelled by the board.
• Scale benchmark: the largest ASX listing since 1997.
• Reason according to the company: continuing is not in shareholders' interests.
• Actual reason: investor demand fell short of expectations.

Why investors got cold feet

AI infrastructure remains one of the most capital-intensive categories in the tech sector. A data center for GPU clusters needs land, a grid connection, cooling, and long-term power contracts. It also needs chips, which are getting more expensive and going obsolete quickly. Revenue depends on how many tenants sign long-term contracts and at what rates. A public-market investor in this model is buying not only current earnings but also a belief that occupancy and margins will hold up through several hardware refresh cycles.

When a valuation bakes in a very optimistic scenario, a small doubt about the customer base, the cost of capital, or the timeline for bringing capacity online is enough to stall the order book. In Firmus's case the details of those doubts have not been disclosed. But the cancellation itself shows that the argument "it's AI" has stopped working as a standalone valuation driver, at least in the Australian market.

There is also a local factor. The ASX is noticeably smaller than the US exchanges, and the pool of funds willing to take a large, high-execution-risk tech offering is limited. A deal of this size needs broad institutional demand. If that demand does not materialize, cutting the price at debut is often worse than walking away.

Practical takeaways

For AI infrastructure players, the signal is mixed. Demand for compute has not gone away, but the public market has become stricter about contract structure, debt load, and the timing of site launches. Companies preparing a listing will have to show up with signed customer contracts and a clear payback model, not just the story of an industry's growth.

For investors, it is a reminder of the difference between a theme and a business. A data center's revenue is determined by rack utilization, not by headlines. Finally, Firmus is worth watching going forward: the cancelled IPO does not mean the end of the project. The company may return to private funding, bring in a strategic partner, or try again under different market conditions. Which path it chooses will show how deep the market's doubt runs.