SUNDAY, 18 OCTOBER 2026GLOBAL AFFAIRS · LONDON
Finance

Firmus IPO collapse shows AI investors are demanding proof, not just scale

The withdrawal of a planned multibillion-dollar Australian listing marks a new test of whether ambitious AI-infrastructure forecasts can withstand public-market scrutiny.

Firmus Technologies has withdrawn its application to list on the Australian Securities Exchange after weak investor demand, abandoning what would have been one of Australia’s largest public offerings.

The Nvidia-backed operator builds modular data centres for artificial-intelligence computing. Firmus said it would instead seek private capital and consider alternative international public-market options.

Why the flotation failed

The proposed offering sought roughly A$7 billion at an initial price of A$11 a share, implying an equity value of about A$44 billion. That valuation had risen rapidly from the company’s previous private funding round and depended on a large expansion programme being delivered successfully.

Prospective investors questioned the price, the company’s expected debt, its limited record of building projects at the proposed scale and uncertainty surrounding a previously discussed development partnership. Reuters reported that Firmus currently has two leased operational sites, in Melbourne and Singapore, and plans five more across the Asia-Pacific region.

The company’s draft prospectus projected annual earnings of about A$5 billion within five years. Public investors were therefore being asked to value a substantial future build-out before much of the capacity was operating.

A wider message for AI finance

The failed deal does not show that demand for AI computing has disappeared. Large technology companies continue to spend heavily on chips, electricity, cooling and data-centre capacity. It does show that investors are separating demand for AI services from the financial quality of every company trying to supply them.

That distinction matters because data centres require large upfront commitments and face execution risks in construction, power access, customer concentration and financing. Revenue may grow quickly, but the return to shareholders depends on funding costs, utilisation, contract quality and how much additional capital is needed before projects produce cash.

Firmus also exposes the difference between private and public markets. Private investors may accept longer time horizons and concentrated risk. Listed-market investors must consider daily liquidity, governance, disclosure, the potential sale of existing shareholders’ stakes and comparisons with established companies.

What comes next

Firmus can still fund expansion privately or attempt another listing at a different valuation or venue. The company’s assets, backers and exposure to AI demand remain valuable. But any renewed financing effort will probably face closer scrutiny of contracts, debt, build schedules and the path from capacity announcements to sustainable earnings.

For the broader market, the lesson is not that the AI investment cycle has ended. It is that capital is becoming more discriminating. Projects with credible customers, secured power, disciplined leverage and visible cash flows should still attract funding. Those relying mainly on scarcity narratives or distant forecasts may find that the cost of capital has risen.

Sources and further reading