October 8, 2026, (Inside AI) — The highly anticipated public listing of Australian data center operator Firmus Technologies is teetering on the edge of collapse. The Nvidia and Blackstone-backed company has been forced to slash the valuation target for its initial public offering by 25%, a dramatic concession that signals growing investor skepticism toward the artificial intelligence infrastructure boom.
Firmus, which is seeking to list on the Australian Securities Exchange, originally pitched a valuation of A$44 billion ($31 billion). According to sources familiar with the deal, the company has now cut the price per share to as low as A$8 from an initial A$11, attempting to hold the line at A$8.25. This revision reduces the total amount the offer is expected to raise to A$5.9 billion ($4.1 billion) from A$7.9 billion, effectively wiping billions off the company's implied equity value.
The struggle to secure backing for one of the largest tech listings in Australian history highlights a widening rift between the narrative of endless AI growth and the cold reality of capital markets. Firmus, a seven-year-old firm co-run by cousins Tim Rosenfield and Oliver Curtis, represents a new class of "neocloud" providers. These companies build specialized, high-density data centers designed specifically to handle the massive computing loads required by generative AI models.
The company's pitch relies heavily on its pivot from domestic partnerships to international expansion. Firmus has shifted its strategy to focus on building capacity in Malaysia and Indonesia to serve hyperscalers like OpenAI and Meta. These contracts are the backbone of its revenue projections, which underpin the aggressive valuation targets. However, the market's appetite for such risk has cooled significantly since the company's last private funding round in August.
A New Metric Fails To Convince Investors
To bridge the gap between traditional financial modeling and the unique economics of data centers, Firmus introduced a bespoke valuation metric known as "EV+1/EBIT+2." The tool was designed to capture the peculiarities of the industry, where companies typically borrow up to 90% of the cost to build facilities and ramp up leverage as construction nears completion.
The metric calculates enterprise value by adding current market capitalization to net debt projected 12 months out, rather than current debt. The denominator, EBIT+2, refers to the earnings before interest and taxes expected two years from now. The logic assumes that data center construction takes roughly 12 months to begin generating revenue, especially in Asian markets, and that hyperscaler contracts make those future earnings highly predictable.
While potentially useful for sectors like mining or biotech, the metric has failed to assuage concerns about Firmus specifically. The company has currently built only 42 megawatts of capacity against a pipeline of roughly 1 gigawatt. This massive gap between current operational reality and future ambition leaves significant room for error.
Critics argue that the EV+1/EBIT+2 model is easily manipulated. Capital expenditure for committed projects can be inflated to pump up enterprise value, and the projected earnings rely on the assumption that demand for AI compute will remain insatiable. If hyperscalers delay their rollouts or if construction faces permitting, equipment, or energy delays, the timeline shifts. In that scenario, EV+1 becomes EV+2 or +3, and the earnings remain elusive.
The skepticism surrounding Firmus is not isolated. It reflects a broader anxiety in the tech sector regarding the sustainability of the AI trade. Investors are increasingly wary of companies that promise future returns based on speculative demand rather than current cash flow. The involvement of Nvidia and Blackstone, while prestigious, does not immunize the firm against market gravity.
The company is allocating around half of its IPO to existing investors, a move that suggests a scramble to ensure the listing proceeds. The shares are expected to start trading on October 23. If the IPO prices at the lower end of the revised range, it will serve as a bellwether for other AI infrastructure startups hoping to tap public markets.
For now, the fate of Firmus remains uncertain. The company is betting that the long-term demand for AI compute will outweigh short-term market jitters. The coming weeks will reveal whether investors share that vision or if the correction in AI valuations has only just begun.
