Firmus IPO puts A$43bn valuation against 900MW of contracted AI capacity

Firmus Technologies is preparing to ask Australian public-market investors to make one of the biggest bets yet on the economics of artificial intelligence infrastructure, pricing its planned initial public offering (IPO) at A$11 a share and targeting approximately A$7.1 billion in proceeds. The terms imply an equity valuation of about US$30.6 billion, or roughly A$43 billion at recent exchange rates, while trading on the Australian Securities Exchange is expected to begin on October 23. The valuation is being supported by an unusually powerful combination of contracted AI-compute demand, more than 900 megawatts of contracted capacity and relationships with Meta Platforms, OpenAI, NVIDIA and Blackstone. The harder question is how much of that future platform investors should capitalise today when two facilities are operating and another five remain under development.

That distinction between contracted demand and operating infrastructure sits at the centre of the Firmus IPO. Firmus has moved beyond being merely a data-centre development concept: customers have committed to capacity, Blackstone and Coatue have backed a US$10 billion financing facility, strategic investors have supplied billions of dollars of equity, and construction is progressing across several locations. At the same time, the IPO is arriving before the market has a long public financial history with which to test utilisation, margins, capital intensity and returns through a complete investment cycle. The result is an unusual proposition in which commercial validation appears to be running considerably ahead of operating maturity.

How did Firmus Technologies reach an implied A$43 billion valuation so quickly?

The speed of the valuation change is one of the most consequential elements of the Firmus IPO. NAOS Ex-50 Opportunities Company, an investor in Firmus, disclosed that an April 2026 US$505 million funding round led by Coatue valued the company at approximately US$5.5 billion post-money. Firmus subsequently announced a fully subscribed US$2 billion strategic equity investment in August involving Coatue, NVIDIA, funds managed by Blackstone and Jane Street, although Firmus did not disclose the valuation attached to that round.

Reuters reported that the August transaction valued Firmus at roughly US$10.5 billion. Against the approximately US$30.6 billion equity valuation implied by the A$11 IPO price, that represents almost a threefold increase in roughly two months and more than five times the April post-money valuation. The comparison does not by itself establish that the new valuation is excessive because the underlying business changed materially during the period, including new customer agreements and an expansion of contracted capacity. It does, however, mean investors are being asked to capitalise those developments extremely quickly rather than waiting for several years of operating evidence.

The timing of those commercial developments helps explain why the valuation trajectory is so steep. On September 8, Firmus said its contracted capacity across customers had surpassed 900 MW after OpenAI agreed to take dedicated compute capacity from two planned Malaysian AI factories. Less than three weeks later, Firmus and Meta Platforms announced agreements covering GPU compute capacity at Firmus facilities in Southeast Asia, while Firmus said Meta was already consuming NVIDIA GB300 NVL72 capacity at its Melbourne operation.

The challenge for investors is that contracting activity and earnings recognition do not occur simultaneously. Long-duration commitments can materially reduce demand risk, but the associated facilities must still be financed, constructed, energised, fitted with expensive computing systems, connected to networks and brought into production before the underlying capacity can generate its expected economics. Firmus is effectively asking the market to value both what it operates and what it has contracted itself to build.

What does more than 900MW of contracted capacity actually tell Firmus investors?

Firmus said in September that its portfolio comprised seven AI factories across Australia, Singapore, Indonesia and Malaysia, with two sites operating and five under development. The company expects the development sites to reach ready-for-service status over approximately 24 months. More than 900 MW of contracted customer capacity therefore provides significant evidence that Firmus is not constructing purely speculative facilities, but it should not be confused with 900 MW of presently operational or revenue-producing capacity.

A useful comparison is NEXTDC Limited, one of Australia’s established listed data-centre operators. NEXTDC reported 740.1 MW of contracted utilisation for fiscal 2026 but only 175 MW of billing utilisation, demonstrating how large the gap can become between capacity sold under contracts and capacity already contributing to current revenue. NEXTDC also reported A$405 million of fiscal 2026 net revenue and A$248.8 million of EBITDA, giving public investors an operating base against which its future contracted megawatts can be assessed.

At its October 2 closing price of A$10.74, NEXTDC had a market capitalisation of approximately A$8.16 billion. Firmus’s implied equity valuation of roughly A$43 billion would therefore be more than five times NEXTDC’s current market value, while Firmus’s disclosed contracted capacity of more than 900 MW is only about 1.2 times NEXTDC’s 740.1 MW. That is not an apples-to-apples valuation comparison because Firmus is positioning itself as an integrated AI-compute infrastructure provider rather than simply a conventional colocation operator, but the size of the gap illustrates how much value the IPO assigns to Firmus’s computing layer, customer relationships, technology platform and expected future margins.

The public-market backdrop also provides a useful caution against assuming that strong data-centre demand automatically protects valuations. NEXTDC shares finished October 2 at A$10.74 compared with a 52-week high of A$17.24, leaving the stock roughly 38% below that peak even while the company reports record contracted utilisation. The share performance does not predict how Firmus will trade, but it demonstrates that public investors can distinguish between long-term infrastructure demand and the valuation they are prepared to attach to that demand at any particular moment.

Why do Meta Platforms, OpenAI, NVIDIA and Blackstone materially strengthen the Firmus case?

Firmus has accumulated a collection of counterparties that substantially lowers one of the biggest risks facing a young infrastructure platform: whether customers will actually arrive. OpenAI has agreed to take dedicated capacity from two Malaysian locations, Meta Platforms has entered agreements covering Southeast Asian capacity, and Firmus says Meta is already using its Melbourne AI infrastructure. NVIDIA, meanwhile, has publicly identified Firmus among the Australian partners building AI factories using its infrastructure ecosystem.

Firmus also announced a strategic compute partnership with NVIDIA extending through 2034 and centred on a planned 360 MW campus in Batam, Indonesia, designed for as many as 170,000 NVIDIA accelerators. The structure is significant because Firmus said the arrangement incorporates economic alignment through revenue sharing and credit support for AI-native customers, potentially reducing some of the financing friction that normally exists between rapidly growing AI companies and capital-intensive infrastructure providers.

Financing support is similarly substantial. Firmus announced in February that Blackstone-led funds and Coatue had backed a US$10 billion debt financing facility intended to support Project Southgate, while the August US$2 billion equity financing brought additional participation from NVIDIA, Blackstone and Jane Street. These transactions do not eliminate construction or utilisation risk, but they show that sophisticated infrastructure and technology investors have been prepared to commit significant capital to the platform.

NAOS Ex-50 Opportunities Company also said in its 2026 annual report that Firmus had approximately A$600 million of annual revenue associated with a hyperscaler contract, alongside the financing and NVIDIA relationship. Because NAOS is itself an investor in Firmus, its characterisation should be read in that context, but the disclosure supplies another indication that parts of the platform have advanced beyond speculative capacity announcements.

Why could power supply and construction execution matter more than Firmus customer announcements?

AI infrastructure increasingly behaves like a heavy industrial project rather than a conventional software expansion. Customer demand can be secured with contracts, but megawatts still require substations, grid connections, transformers, cooling equipment, buildings, fibre, backup systems and enormous quantities of computing hardware. Firmus has therefore been attempting to control more of the physical chain rather than relying entirely on third-party construction and infrastructure providers.

The company agreed in August to acquire the fabrication, design and projects operations of Benmax for A$300 million, saying the transaction would bring design, fabrication and construction deeper into its vertically integrated AI Factory model. Earlier, Firmus announced a 12-year wholesale electricity agreement with Gunvor for 600 MW of firm supply in South Australia, linked to the development of 1.2 GW of new renewable generation and 1.5 GWh of battery storage by 2032. Those agreements provide evidence that the company is working on the physical inputs required to turn customer commitments into operating compute rather than treating power as somebody else’s problem.

There is also evidence of actual construction progress. MAAS Group Holdings disclosed in May that its A$200 million contract for Firmus’s Launceston AI Factory was approximately 35% complete and remained on track for calendar-2026 commissioning. Firmus currently describes its Southgate Tasmania operation around an 84 MW critical IT load, while broader Australian and regional expansion plans extend far beyond that initial footprint.

This is precisely where the investment proposition moves from customer validation to execution mathematics. A long-term contract may reduce demand uncertainty, but delayed energisation can postpone revenue while interest, engineering and procurement costs continue. Conversely, bringing capacity online faster than competitors could allow Firmus to monetise a period in which scarce AI compute remains strategically valuable. The IPO therefore depends not merely on whether artificial intelligence demand grows, but on whether Firmus can convert contracted megawatts into billable megawatts on schedule and at the capital cost embedded in its valuation assumptions.

What does the reported US$30 billion debt estimate mean for the Firmus IPO?

The capital structure may become one of the most closely examined sections of the formal prospectus. Reuters reported that analysts working for banks involved in the IPO estimated Firmus would have approximately US$30 billion of debt, which together with the implied US$30.6 billion equity valuation would produce an enterprise value of roughly US$60 billion. That figure is considerably larger than the US$10 billion financing facility Firmus publicly announced in February.

Those figures should not automatically be treated as contradictory. A committed financing facility is not necessarily equivalent to drawn debt on a balance sheet, while a bank estimate of future indebtedness may incorporate additional project financing, equipment financing or capital requirements associated with the expansion programme. Until the public prospectus provides the detailed capital structure, maturities, interest costs, drawn amounts and funding assumptions, treating US$30 billion as straightforward current balance-sheet debt would risk overstating what is presently known.

That distinction matters enormously for equity investors. Infrastructure businesses can support substantial leverage when contracts are long-lived, counterparties are strong and cash flows are predictable, but a heavily leveraged development programme becomes much more sensitive to delays, capital-cost overruns, customer concentration and financing costs. An A$7.1 billion IPO would give Firmus another large pool of equity capital, but the investment case still requires investors to understand exactly how much additional capital will be required before the five development sites reach productive operation.

What must the Firmus prospectus prove before the October 23 ASX listing?

The most important document in the Firmus IPO has not yet entered the public record. Reuters reported that the prospectus is expected to become public on October 8, following the start of institutional bookbuilding, meaning several headline figures currently circulating are derived from term sheets, investor materials, company announcements and reporting on draft documents rather than a final statutory offer document.

The prospectus will therefore need to connect several pieces of the story that currently exist separately. Investors will be able to examine how contracted capacity converts into expected revenue, how much of the customer portfolio represents binding commitments rather than expansion options, the economics of GPU ownership and procurement, the timing of capital expenditure, the treatment of project debt and the margins expected once each facility reaches steady-state utilisation. It should also make clearer how Firmus’s integrated AI-compute model differs economically from traditional data-centre companies, because that distinction is central to any justification for the valuation premium.

Pre-IPO sentiment already appears divided rather than uniformly euphoric. Reuters reported that indicative demand exceeded the targeted deal size even as some Australian fund managers questioned whether the company’s short operating history and valuation could support the expectations embedded in the offer. Those positions are not mutually exclusive: an IPO can be heavily subscribed while still producing sharp disagreement about long-term value, particularly when scarcity, AI exposure and a high-profile customer list attract investors with very different time horizons.

Firmus has already answered one of the hardest questions facing an emerging infrastructure company by demonstrating that large technology customers want what it is building. The October listing will test a different proposition: whether public investors are prepared to pay today for the economics of infrastructure that is still being constructed. If Firmus converts more than 900 MW of contracted demand into productive AI factories on schedule and at attractive returns, its valuation may increasingly be judged against global compute platforms rather than conventional Australian data centres. If execution, financing or utilisation falls behind, the enormous amount of future performance already embedded in the IPO price could make that same valuation considerably less forgiving.