You’ve been warned

It would just be helpful if they were priced in a way that doesn’t almost guarantee IPO investors a long-term loss.

I’m referring to the upcoming IPO of Firmus, Australia’s contribution to the AI bubble.

That its prospectus – for Australia’s second-largest IPO ever – is not publicly available yet tells you all you need to know. (It’s reportedly due on 12 October.)

That is, the more hype and FOMO that can be created, and the less actual information there is to temper the FOMO, the better.

The brokers and bankers are about to hit the poor old ‘retail’ investor, the credulous souls who want to believe, with an offer to buy shares.

The key to an absurd price for any stock is the ability of promoters to get the imaginative juices flowing.

Firmus isn’t just any old data centre, thanks very much. It’s building ‘AI Factories’, filled with the latest Nvidia chips. Its competitive advantage is some whizz-bang cooling technology that allows it to reduce the cost of producing tokens. (Tokens are the currency of AI usage.)

It would be nice to think the Firmus listing is the culmination of a lifelong pursuit by the founders.

But it was formed in 2019 as a Bitcoin miner…because Bitcoin was all the rage back then. When Bitcoin went into one of its winters, the enthusiasm for the digital currency disappeared.

Out of that disappointment grew a belief in the next hot thing…AI!

And here we (or they) are, convincing everyone that a business with just two functioning data centres plus lots of proposed ‘AI factories’ has an equity value of A$44 billion, plus potential debt of US$30 billion (or A$43 billion)…but we don’t really know because the prospectus isn’t available yet.

Regular data centres rent out space to customers, and customers bring their own servers and chips.

Firmus will build AI factories – called hypercubes – fill them with Nvidia chips, advanced liquid cooling, and networking systems, and rent the ‘full stack’ to customers. It says it can do this cheaper than anyone else, meaning lower factory running costs and lower token costs.

That none of this is proven doesn’t seem to bother anyone.

The business is impossible to value accurately. But we can at least ask whether a $44 billion value at the IPO price is reasonable.

A share price can go up or down for many different reasons, mostly relating to the psychological state of investors trading the stock that day.

But a business (which is what a share price represents at the end of the day) creates value when its assets generate a greater return than the cost of the capital that funds them.

The balance sheet shows a company’s assets and how it funds them (debt plus equity). Without the prospectus, it is impossible to know what the balance sheet looks like.

But we can make a rough guess. Estimates of debt are US$30 billion, or around A$43 billion. The balance sheet equity value could be around A$12 billion, which is an estimate of the cumulative value of the pre-IPO and IPO equity raisings.

That gives balance sheet assets of approximately A$55 billion: A$43 billion in debt and A$12 billion in equity (roughly 80% debt, 20% equity).

Based on my research, the cost of debt financing looks to be around the 10% mark. Let’s assume the cost of equity for a rational investor is 15%.

Blended, that’s a weighted cost of capital of 11%. (Nerd note: I use after-tax cost of debt and market value of equity to arrive at this estimate.)

So, let’s assume it costs Firmus 11% to fund its assets. That means, to earn its cost of capital, Firmus needs to generate net operating profit after tax (NOPAT) of $6.05 billion.

Firmus is ‘targeting’ earnings before interest and tax of at least US$5 billion in 2028, or A$7.2 billion.

Allow for tax (and for the fact that more capital will be invested by then) and it’s a long shot that Firmus will be earning its cost of capital in a few years’ time.

But here’s the important point…

Companies that earn below their cost of capital generally trade at a discount to their equity book value, if they’re being valued rationally. Firmus’ book value, on my rough estimates, is $12 billion. Yet the IPO price puts the equity value at $44 billion, 3.67 times book value!

No wonder this deal needs lots of imagination.

Look, it’s not unusual for a start-up to earn below its cost of capital for a few years. As long as the revenue line is increasing strongly, markets will often see through to a period when profits and profitability start rising strongly.

But in this case, Firmus will need to keep spending. That’s why it’s listing. It needs access to debt and equity markets and enough credulous investors to keep the money flowing.

But if it’s not even covering its cost of capital in a few years’ time, with plenty more capital needed to continue growing, the IPO valuation is absurd.

This view may not show up in the share price right away. But I don’t think it will take long for the market to realise what a capital-intensive beast this is. The equity value (the share price) will start to fall sharply when punters realise they won’t be getting a return on their investment for a long time.

Debt investors owed US$30 billion get paid first.

You only need to look to the Nasdaq-listed CoreWeave, a company Firmus compares itself to, to see how capital-hungry, and devoid of returns on this capital, the business is.

Consensus estimates point to US$42.5 billion in revenue in 2028, and just US$740 million in net profit. That’s a tiny profit margin.

Free cashflow is expected to be NEGATIVE US$21.5 billion in 2028.

That’s why CoreWeave’s debt has increased from US$1.3 billion in December 2023 to US$35 billion as at 30 June. And that doesn’t include lease liabilities of US$16 billion.

By comparing Firmus with CoreWeave, I’m guessing the promoters are assuming no one will look at CoreWeave’s accounts.

It all looks like a ticking time bomb to me. Maybe I just don’t get it. It wouldn’t be the first time.

Forget the equity value – I’d be surprised if the debt holders get out of this arrangement with 100 cents on the dollar.

At some point, the tide is going to go out on this AI infrastructure build-out. It will reveal A LOT of people have been swimming naked. The credit funds supplying the debt will be in trouble.

If Firmus can get its IPO away successfully and have enough punters buying in the market to keep the price up for a few months, it will be one of the greatest sales jobs in Australian capital markets history.

For anyone who loses money on this, there can be no complaining. It’s right in front of you, plain as day.

As I said, sometimes people just want to believe, despite all the evidence to the contrary.