Firmus $44b ASX IPO collapse exposes investor doubts over Nvidia-backed AI data centre giant’s ambitious plans

The collapse of Firmus’ share market listing was less a verdict on greedy investment bankers than on management’s failure to convince investors it could deliver one of the most ambitious data centre construction programs in the world.

Initial public offering (IPO) investors rejected the chance to buy shares at a $US30.4 billion ($43.7 billion) valuation, which investment bankers actually pitched at a cheap-looking 5.2 forecast earnings before interest and tax (EBIT) of $US5.8 billion in calendar year 2029.

For a fast-growing AI infrastructure business, this isn’t expensive by any conventional valuation metric.

On the contrary, Firmus shares were offered pretty cheap. The problem wasn’t the valuation or greedy, incompetent investment bankers.

It was that professional investors weren’t persuaded by the Firmus management team, its plans, or forecasts to go from losses today to $US5.8 billion in profits by 2029.

The bankers even cut the IPO price from $11 a share at a $43.7 billion valuation to $9 a share, then $8, and even half price $5.50 - equal to just 2.6 times forecast profits in 2029.

But there were still not enough large buyers of shares to believe the forecasts of a company that would’ve been the largest IPO since Telstra in 1997.

In reality, if Firmus delivers on its forecast EBIT of $US5.8 billion by 2029, Australian investors have just made a historic and expensive mistake.

Worse, Firmus could yet list on the Nasdaq and become a spectacular success, proving an AI wealth windfall for overseas investors.

Firmus’ prospectus even points to independent research that forecasts demand for AI compute is set to multiply a staggering 162-fold - between 2025 and 2030.

Its prospectus implied it offered a way to go to where the puck is heading and invest in that kind of vertical growth only linked to AI via an ASX-listed stock.

Its backers include Nvidia and Blackstone, with OpenAI and Meta among its customers. The US financiers, including Nvidia, were even prepared to fund most of Firmus’ construction projects via debt and equity investments.

ASX investors were being asked to contribute a relatively small amount of equity along the way on a modest valuation if the earnings forecasts were to be believed.

That should have been an attractive proposition and should’ve made a compelling case for the shares to be worth considerably more than $U11 by 2027 or 2028.

But the market’s verdict was clear. Firmus’ management hadn’t established that its forecasts and timetable to build and operate five new data centres by the end of 2028 were credible.

Moreover, Firmus’ forecasts for exceptionally low costs were especially hard to believe.

Credibility gap

The IPO exposed a yawning gap between the scale of the opportunity presented on paper and hard evidence that Firmus’ management could capture it to budget in the real world.

One of the biggest stumbling blocks was the company’s claim that it could build data centres for roughly half the cost of established competitors.

Firmus attributed this shock advantage to its proprietary ‘Hypercube’ technology. But investors were being asked to accept a substantial cost advantage without enough evidence that the tech could deliver the promised savings at scale.

“We were avoiding [Firmus] at any price,” said Ron Shamgar, the founder of Tamim Asset Management.

The limited prospectus also revealed a concentration of forecast profits in Malaysia and Indonesia, which raised questions and eyebrows about jurisdictional risk.

About half the forecast EBIT of $US5.8 billion was expected to come from a data centre under construction in Batam, Indonesia, with another 32 per cent generated by two Malaysian data centres.

That left around 82 per cent of the forecast profit dependent on the successful construction and operation of infrastructure in two countries with risks around local regulation, corporate governance, politics, environmental approvals, water availability, power supply and construction costs.

The obstacles were real for a company whose valuation depended on delivering huge amounts of compute capacity to tight timelines without hitches or cost blowouts.

How much debt?

The lack of detail and disclosure around the funding plan was another weakness.

Firmus’ prospectus forecast capital expenditures of $US37.7 billion to build its five new data centres, including the high cost of buying computer chips from Nvidia, itself an investor in the company.

Yet the market was left with limited clarity about how that forecast expenditure would be financed and why it was reliable.

A rough expectation of $US7 billion in equity funding, largely via the failed IPO and $US30 billion in debt from mega backers made sense.

But the company refused to answer basic questions about how it would fund its spending plans, which remain largely a secret. The prospectus mentioned a $US10 billion debt facility, but gave scant detail.

The lack of disclosure about its future funding plans also meant investors were left grappling with the potential for further share issuance if the company needed to raise additional cash and dilute existing shareholders.

Inadequate restrictions on existing shareholders selling immediately after the listing were also flagged as major worries for professional investors, who did not want the shares to quickly lose value if early investors had too much capacity to sell out.

Terms disclosed in the prospectus also restricted dividend payments to equity investors as effective claims on the future profits until the debt holders were repaid.

In effect the equity investors were being sold ‘trust us’ vibes by Firmus’ management, rather than evidence the company could build its infrastructure on budget and turn contracted demand into sustainable profits.

Elsewhere, selling hopes and dreams around data centres in space to a brainless crowd of retail investors has worked for Elon Musk at SpaceX.

But Firmus’ co-founder Oliver Curtis is no Elon Musk. He also had to sell his dream to a sophisticated crowd of professional investors, rather than the millions of less-sophisticated investors guided by social media to buy SpaceX.

Wrong timing

Firmus was also under pressure to list quickly under the terms of funding agreements with Nvidia.

The race to secure more capital as tech companies scramble to build AI infrastructure may have pushed the company to market before its investment case was ready.

According to its prospectus, Firmus expected to have seven data centres operating and billing customers by the end of 2028.

Just another 12 months of operating execution could have shown construction was progressing on schedule, costs were under control, customers were taking capacity, and revenue was beginning to flow towards its stated value of $US72.6 billion over the lifetime of signed contracts with customers like Meta and OpenAI.

The stock would then have looked decent value at just more than five times forecast EBIT in 2029, even if investors still demanded a discount for execution, re-contracting and depreciation risks.

Debt pile

Most professional investors also argue its significant debt pile must be taken into account when valuing its equity.

Still, even after adjusting for the estimated debt amassed of up to $US30 billion to fund the data centres construction and operation, Firmus would still be trading on an enterprise value (debt of $US30 billion + equity market value of $30 billion) around $US60 billion, or less than 12 times its forecast EBIT.

This is significantly cheaper than the current average multiple of 17 times next year’s estimated profits for companies in the S&P/ASX 200, according to Market Index.

Firmus’ forecast for $US5.8 billion in EBIT also incorporates the substantial depreciation costs associated with expensive computing chips and data-centre infrastructure losing value.

So, the failed IPO does not prove Firmus is a bad business, or that its forecasts are impossible.

Nor does it establish that investors were wrong to walk away. It shows that the potential rewards were not enough to compensate them for the uncertainty surrounding the profit forecasts or management’s ability.

The investment bankers could cut the price in half. They could not manufacture the missing confidence in the company.

Firmus made the mistake of asking professional investors to believe its ambitions. It would’ve been better off waiting 12 months to show them what it had built.