Monologue: Australia's Dodgiest AI IPO Is A Pale Horse

A ten-minute Better Offline monologue, aired 9 October 2026, about Firmus, an Australian AI data-centre company backed by Nvidia, and its attempt to float. Zitron calls it a pale horse, the omen before the rest.

His case against the company is the case he makes against every neocloud. Firmus has little revenue and plans to build 10 gigawatts in five years. It has a US$10 billion loan from Blackstone and Coatue, the same backers as CoreWeave, and most of its backlog sits with two customers, Meta and an unnamed technology company, plus a compute deal with OpenAI from Malaysian data centres that don’t yet exist. Its build with CDC, Project Southgate, has been dissolved with about 2.6 per cent of it built. The IPO, once meant to raise about A$7 billion and value the company near A$50 billion, was cut from A$11 a share to A$9 as investors balked. UniSuper said the valuation didn’t add up, and a fund manager at Plato said the price assumed five years of flawless execution.

His real target is the press. Australia’s financial journalists, he says, went after Firmus for months, talking to sceptic after sceptic and questioning the banks, which is what American and British reporters never do for CoreWeave, Nebius, IREN or Lambda, though every criticism applies to them too. His example of the other kind is Bloomberg covering CoreWeave’s 240-megawatt Indian deal with AdaniConneX, not due until mid-2028, as a story about expansion. When this collapses, he says, part of the blame will be the media’s, and the losses will land on pensions and insurance funds. Blackstone’s credit and insurance fund is among Firmus’s backers. By the time he spoke the float was already falling apart: Bloomberg had reported the day before that Firmus would postpone it and look at a private round.