Firmus IPO stumble puts a dent in the neocloud boom

Firmus Technologies wanted public investors to value it at A$44 billion (approximately $31 billion). Investors had other ideas.

The Australian data center developer cut its offering price by 25% on October 7, 2026, according to a Reuters Breakingviews report published the following day. Breakingviews framed the $4 billion neocloud IPO as a pinprick in the AI bubble, and the timing is awkward. Several of Firmus’s peers are lining up to raise billions of their own.

What went wrong with the Firmus listing

Firmus is a neocloud, one of a newer breed of companies built to rent out GPU capacity for AI workloads. They build or lease data centers, pack them with high-end chips, and sign long-term contracts with hyperscalers that need the horsepower.

Firmus’s own numbers show why execution uncertainty matters. The company has just 42 MW of operational capacity. Its development pipeline sits at around 1 GW.

Then there is the valuation math. Firmus used a novel metric, labeled “EV+1/EBIT+2,” to help justify its price tag. That target was approximately triple the valuation from its most recent funding round in August 2026.

A strategic pivot and a complicated backstory

Firmus is also changing where it plans to grow. The company is shifting its geographic focus toward Malaysia and Indonesia as it seeks contracts with major clients such as OpenAI and Meta. That move pulls it away from a domestic partnership with CDC Data Centres.

The cap table, at least, looks impressive. Firmus counts Nvidia and Blackstone among its backers.

Leadership is a different conversation. The company was co-founded by Tim Rosenfield and Oliver Curtis, and Curtis has a prior conviction for insider trading.

A 25% price cut is not a rounding error. It reflects a meaningful gap between what Firmus thought it was worth and what buyers were willing to pay.

The neocloud queue gets longer

Firmus is not raising money in isolation. Nscale is targeting a listing on the NYSE, with an anticipated raise of $3 to $3.36 billion. Lambda is aiming for a $4 billion pre-IPO raise at a valuation of around $14.5 billion.

What this means for AI infrastructure

Breakingviews’ read suggests investor hesitation around Firmus could bring more scrutiny to other neocloud listings.

The specific weak points in the Firmus pitch are worth watching across the sector. The first is the ratio of built capacity to planned capacity. A company with 42 MW running and around 1 GW planned is asking investors to underwrite a lot of future construction.

The second is customer concentration. Firmus’s pursuit of clients like OpenAI and Meta in new markets highlights how much rides on a few relationships.

The third is valuation discipline. When a company needs a brand new metric to explain a valuation roughly three times its last private round, investors are likely to ask why the standard measures were not good enough.