Nvidia-Backed Firmus Lands $505M at $5.5B Valuation to Expand Asia-Pacific AI Data Centers

Image: Bloomberg AI
Main Takeaway
Firmus secures $505M led by Coatue, bringing total raised to $1.35B in six months as it races to build AI infrastructure across Asia.
Summary
The funding round and valuation
Firmus Technologies Pty has closed a $505 million strategic equity round led by Coatue Management at a post-money valuation of $5.5 billion, according to announcements from the Singapore-headquartered data center company on Monday. This marks the second major capital injection for the Australian-founded firm in barely half a year, pushing total fundraising to $1.35 billion since October as the AI infrastructure arms race accelerates across the Asia-Pacific region. The deal cements Firmus as one of the highest-valued private builders of AI-optimized facilities in the southern hemisphere and signals continued investor appetite for physical compute assets rather than just software plays.
Nvidia's strategic involvement and technical backing
Nvidia’s participation goes beyond a passive investment. The chip giant is providing reference architectures, supply guarantees for its latest H200 and upcoming B100 GPUs, and early access to its liquid-cooling toolkits, according to people familiar with the partnership. That backing lets Firmus promise tenants a direct pipeline to the newest hardware without the eighteen-month wait most cloud providers face. The company brands its facilities as “Southgate” data centers, a nod to both their southern-hemisphere footprint and the gateway role they play between Nvidia’s silicon roadmap and regional hyperscalers. Analysts note that Nvidia’s co-marketing gives Firmus an edge similar to the one OpenAI gained from Microsoft’s Azure credits, except applied to bricks and megawatts instead of cloud credits.
Expansion plans across Asia-Pacific markets
Proceeds from the round will bankroll a four-country build-out: new campuses in Sydney and Melbourne, a 48-megawatt site in Johor Bahru that straddles the Singapore border, and two speculative developments in Tokyo’s Tama district. Together these projects add roughly 300 megawatts of critical IT load by 2028, enough to house roughly 120,000 high-end GPUs at standard density. The Johor site is particularly strategic; it sits on a fiber landing station that gives Singapore-based banks and trading houses sub-millisecond latency while sidestepping the city-state’s land-scarcity constraints. Meanwhile, Japan’s land prices and stable grid make it attractive for long-term training clusters, an angle Firmus is pitching to domestic conglomerates that need sovereign AI capacity.
Competitive landscape and investor rationale
Coatue’s check reflects a broader rotation among growth investors from AI model builders to the picks-and-shovels layer after last year’s software valuation reset. While Amazon, Microsoft, and Google still dominate cloud GPUs, regional players like Firmus, Lamda, and Yotta are carving off niches with faster deployment, sovereign data guarantees, and access to cheaper renewable power. Coatue partner Rahul Kothari told investors the firm sees data centers as “the new oil refineries” of the AI age, with scarcity rents lasting at least through 2030. The $5.5 billion valuation implies roughly 22× forward EBITDA based on contracted power, richer than traditional colocation REITs but below the 30-35× multiples fetched by edge-data-center peers Digital Realty and Equinix during their own growth spurts.
What happens next for AI infrastructure investors
Firmus has already filed preliminary paperwork for a potential 2027 IPO on the ASX, contingent on securing another $725 million in project-finance debt later this year, according to sources close to the process. That would make it the first pure-play AI data-center operator to list in Australia and could open the door for similar regional specialists to tap public markets. More immediately, the company must lock in long-term power-purchase agreements amid soaring electricity prices in New South Wales and navigate Malaysia’s shifting foreign-ownership caps on critical infrastructure. Success or failure will serve as a bellwether for whether the current wave of AI infrastructure valuations is grounded in real cash-flow capacity or simply momentum chasing the next shiny object.
Key Points
Firmus closed a $505M strategic round led by Coatue, pushing its six-month fundraising total to $1.35B and valuing the company at $5.5B.
Nvidia provides more than capital: priority GPU allocation, liquid-cooling IP, and co-marketing underpin the Southgate data-center brand.
Expansion roadmap covers 300 MW of new IT load across Australia, Malaysia, and Japan by 2028, targeting sovereign AI and low-latency trading workloads.
Investor thesis treats data centers as the new oil refineries, with scarcity rents lasting through 2030 amid regional power and land constraints.
IPO filing is already underway for 2027 ASX debut, contingent on an additional $725M project-finance raise and regulatory clearances.
FAQs
Firmus constructs and operates high-density AI data centers—branded Southgate facilities—optimized for Nvidia’s latest GPUs, featuring advanced liquid cooling, high-rack power density, and direct interconnection to major cloud and trading networks.
By guaranteeing priority GPU supply and sharing reference designs, Nvidia ensures its hardware gets deployed quickly in underserved regions, locking in long-term demand and preventing customers from defecting to rival chip vendors.
The facility sits just across the border in Malaysia on a fiber landing station, giving Singapore-based users sub-millisecond latency while accessing cheaper land and power outside the city-state’s constrained real-estate market.
Key risks include soaring electricity prices in Australia, Malaysia’s evolving foreign-ownership caps on critical infrastructure, and the possibility that AI demand forecasts prove overly optimistic, compressing contracted-power multiples.
It offers faster deployment (12–18 months versus 3–4 years), sovereign data guarantees for regulated industries, and access to renewable power contracts that hyperscalers sometimes struggle to secure in smaller markets.
Source Reliability
40% of sources are trusted · Avg reliability: 73
Go deeper with Organic Intel
Our AI for Your Business systems give you practical, step-by-step guides based on stories like this.
Explore ai for your business systems