Australian AI infrastructure company Firmus withdrew its planned IPO before a single share began trading, after investors balked at the proposed valuation and the funding demands of its data-centre expansion. The collapse is a sharp reminder that demand for AI computing does not automatically make unfinished infrastructure an easy sell to public investors.
Watch Desk analysis
What happened
Firmus had planned to raise about $4.9bn, with the offer potentially rising to $5.5bn, at a proposed A$11 a share. The offering was seeking a valuation approaching $30bn, nearly three times the roughly $10.5bn valuation from an August financing round. Coinpaper reports that Firmus abandoned the IPO on 9 October, citing Reuters.
The gap investors were being asked to bridge was substantial: Firmus had about 46 megawatts of operating data-centre capacity, against an expansion pipeline exceeding 900 megawatts. Building that capacity would require major spending on Nvidia processors, electricity connections, cooling and construction. Firmus said in its withdrawal statement that the available offering terms did not adequately reflect its business and long-term outlook, according to Coinpaper.
Read Coinpaper’s account of the failed IPO.
Why it matters
AI infrastructure businesses often need to raise money and build facilities before they can sell computing capacity at scale. Firmus’s experience shows the other side of the AI build-out: investors may like the destination, but still question the price, the construction timetable and who will fund the journey.
The IPO withdrawal does not mean Firmus has failed as an operating company, or that its data-centre plans cannot succeed. It means this proposed public-market deal did not attract enough demand on the terms offered. Coinpaper says Firmus is now seeking private financing while considering other possible listings.
Our read
An AI boom can make a $30bn valuation sound like the obvious next line on a chart. But 46MW operating and more than 900MW in the pipeline are very different things. The question for investors is not only how much AI compute might be needed, but how much capital must be spent before that capacity exists and earns revenue. The market has supplied an answer to this offer, not a verdict on the entire business.
What to watch
- Whether Firmus secures private financing, and on what terms.
- Whether the company revises its valuation or timing for a future listing.
- How quickly its planned data-centre capacity moves from pipeline to operating infrastructure. Activity teaser: Firmus’s planned $30bn IPO has collapsed before trading, after investors questioned the valuation and the cost of building out its AI data centres. The company had about 46MW of operating capacity against a pipeline exceeding 900MW, a sizeable gap between what exists and what still needs funding. The failed offer is not the same as a failed business, but it puts a useful question to the AI infrastructure boom: how much should investors pay for capacity that has yet to be built?
Discussion spark: Should investors value AI data-centre companies mainly on the computing demand they expect, or on the infrastructure and financing they can demonstrate today?
Sources and evidence
- Source update (17 October 2026, 18:49 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.