Why Investors Are Becoming More Selective About High-Valuation Technology Companies

Investors have not stopped buying technology. They are asking more questions first. This week's collapse of Firmus's planned Australian listing showed how quickly an AI-linked offering can stall when price, debt and proof of earnings do not line up.

The Firmus case

Firmus, an Nvidia-backed operator of AI data centres, withdrew its application to list on the ASX on Friday, October 9. The offer was priced at A$11 a share. Reuters reported that this implied an equity valuation of about US$30.6 billion, nearly triple the US$10.5 billion set in a funding round in early August. The Bull put the valuation at about A$43.7 billion.

Demand did not hold up. Reuters said investors pulled orders after CDC Data Centres' chief executive said a planned 1.6 GW project with Firmus was no longer going ahead. Reuters also listed concerns about debt, the company's limited record of building AI data centres, and escrow terms that would have let existing holders sell more than half the stock from day one.

Debt was the sharpest issue. ABC News reported that Firmus expects about US$30 billion of debt once its data centres are built, roughly six times the US$5 billion of operating earnings it forecasts for 2028. Only two of its seven planned sites are operating. The Bull reported a projected loss of about A$77 million for the first half of FY2027, a figure I could not confirm elsewhere.

Firmus's board said the terms would not reflect its long-term prospects, and the company will pursue private funding. UniSuper's John Pearce called it "priced to perfection", while Ten Cap's Jun Bei Liu called it a reality check but not the end of the AI trade.

Why price and interest rates matter

A high valuation assumes a lot of earnings far in the future. When interest rates rise, those distant earnings are worth less today, and safer alternatives pay more.

That pressure is visible now. Forbes reported on October 7 that the US 10-year Treasury yield had risen above 5.35%, its highest since 2002. For a company priced on profits that may arrive in five or ten years, that shift matters more than it does for a business already earning cash.

What investors want to see

Revenue growth still counts, but it is no longer enough on its own. Investors are looking at four things:

• Profitability: whether margins improve as the business scales.
• Cash flow: whether operations fund growth or debt does.
• Contracts: whether customers are committed, and for how long.
• Execution: whether the company has built at this scale before.

Morningstar's Lochlan Holloway noted that neo-cloud firms borrow against customer contracts to buy chips. That works while contracts hold. It looks fragile when a partner walks away.

Will the AI spending pay off?

Nobody knows yet, and the answer probably differs by company. FactSet reported in July that the five largest hyperscalers (Alphabet, Amazon, Meta, Microsoft and Oracle) were expected to spend more than US$690 billion on capital expenditure in FY26, up more than 80% on the year. It said debt had grown from 9% of capex in FY24 to 32% on a trailing basis, and that free cash flow was expected to be near zero or negative for all but Alphabet and Microsoft.

Revenue is growing too. A Forbes contributor, Jason Kirsch, cited Google Cloud growth of 63% in the first quarter, but argued spending is still rising faster than sales. FactSet noted that total debt is about one times EBITDA or lower for four of the five companies, with Oracle the outlier. S&P cut Oracle to BBB- in July.

Returns also depend on end demand. Reuters noted US-listed chipmakers are up more than 80% this year, yet fell 3.4% on Thursday after reports that OpenAI's September annualised revenue was about US$50 billion, below earlier signals.

Winners and those facing disruption

Investors are also sorting companies into those that sell the picks and shovels and those whose products AI might replace. In January the iShares software ETF (IGV) fell 15%, its worst month since 2008, as fears grew that AI tools could erode demand for software, Axios reported. Anthropic said its tools complement software providers rather than compete with them.

The sell-off was broad, but the questions that followed were specific: which firms hold proprietary data, entrenched customers and pricing power.

What it could mean for startups and funds

Late-stage startups may face tougher IPO pricing and fewer exits at inflated private valuations. Venture capital may lean toward companies with clear paths to profit. Public-market investors may hold quality AI names while declining to pay any price. Firmus is turning to private capital instead.

None of this proves a bubble. Strong companies still raise money, and Pearce said Firmus has a strong story. The point is that valuation has to be earned.

What investors will check next

Before committing capital, investors are likely to ask five things: how much debt sits behind the growth, how much revenue is contracted rather than hoped for, how soon projects start earning, what happens if a key partner leaves, and whether early holders can sell immediately. Companies that answer clearly can still raise money at healthy prices. Those that cannot may find, as Firmus did, that the market will not wait.