Firmus IPO Collapse and ASIC Inquiry, October 2026

Firmus abandoned its A$7 billion (~US$5 billion) ASX listing after weak demand challenged a roughly US$43 billion valuation, reportedly drawing ASIC's attention, while the company pivots to a private fundraise of up to US$3 billion.

The disclosure that Australia's securities regulator is reportedly in conversation with investment bankers over the collapse of Firmus's initial public offering adds a formal dimension to what had already become one of the most closely watched capital-markets stumbles in Asia-Pacific's AI infrastructure buildout. The Australian Securities and Investments Commission has not opened a formal investigation, according to October 2026 reporting, but the conversations themselves signal that the regulator is scrutinising the circumstances surrounding the abandoned float — and, more broadly, the questions of disclosure and valuation methodology that accompanied Firmus's rapid rise toward the public markets.

The sequence of events is well documented across multiple reports. Firmus, the Nvidia-backed Australian data-centre operator, targeted an A$7 billion (approximately US$5 billion) listing on the ASX that would have ranked as the exchange's second-largest IPO on record. The company was pitching a valuation of roughly US$43–44 billion at the time of roadshowing — a figure that multiple outlets described as heroic, and that by one account had tripled from implied levels in the space of two months. When subscription demand — particularly from US investors, according to Futu — fell short, Firmus cut its target share price, then withdrew the offering entirely and pivoted toward a private fundraise of up to US$3 billion, as reported by the Wall Street Journal, among others.

What makes the reversal striking is the speed and scale of Firmus's ascent before the stumble. The company had converted from Bitcoin mining operations and by August 2026 attracted equity from Coatue, Nvidia, Blackstone, and Jane Street in a US$2 billion round that pushed its implied valuation past US$10.5 billion. By September 2026, it had announced a multi-year compute agreement with OpenAI for AI factory infrastructure in Malaysia, a separate capacity deal with Meta covering Southeast Asia, a connectivity partnership with SUBCO, and a US$300 million arrangement with Slattery centred on Australian AI data-centre development. Contracted capacity reportedly surpassed 900 MW. The Maas Group disclosed an A$855 million construction and power-infrastructure contract with Firmus in August 2026, lifting its own earnings guidance — concrete evidence that capital was moving into the ground.

The roadshow, however, surfaced a tension that has become endemic to the neocloud sector: the gap between contracted-capacity ambitions and near-term revenue visibility. The Guardian noted investor scepticism about Firmus's operational track record, and The Australian flagged debate over whether the valuation comparisons underpinning the IPO were supportable. The allocation of roughly half the IPO to existing investors, as reported, also raised questions about secondary-market liquidity for new institutional buyers. The broader context is unforgiving: industry executives cited in October 2026 reporting now expect only seven to ten neoclouds globally to survive out of the several hundred that have emerged, with Nvidia identified as the entity with the most to lose in a consolidation. AirTrunk's decision to delay its own IPO and PLDT's parallel pullback suggest the demand weakness Firmus encountered reflects a broader cooling in public-market appetite for pre-profitability AI infrastructure at peak valuations, rather than a problem unique to Firmus.

Three signals will determine whether Firmus can rebuild momentum. First, whether it closes a private fundraise of up to US$3 billion at a valuation that credibly bridges the gap between its August 2026 private-round pricing and the abandoned public-market ask — a successful close at a stable price would confirm that institutional demand exists even without a public listing. Second, the pace at which contracted capacity — the 900-MW-plus figure reported as of September 2026 — converts into delivered megawatts and eventually into billings; the OpenAI and Meta agreements are headline-grade partnerships, but their commercial terms and ramp schedules remain undisclosed. Third, the trajectory of ASIC's engagement: if preliminary conversations escalate into a formal investigation, the scrutiny could materially affect fundraising timelines and the credibility of any future listing, whether on the ASX or the Nasdaq that Firmus had reportedly flagged as a longer-term destination. The structural case for AI compute in Asia-Pacific remains intact, and Firmus's contracted pipeline is not negligible — but its experience underscores that the distance between infrastructure ambition and investor-grade financial transparency is wider, and harder to close quickly, than its valuation trajectory had implied.