Startups

Firmus Raises $2B at $10.5B Valuation for AI Data Centers

The former Bitcoin miner doubled its valuation in four months as institutional investors treat GPU infrastructure like power plants.

Omega Editorial· August 9, 2026· 3 min read

Institutional capital treats AI infrastructure as hard assets

Firmus, an Australian company that pivoted from Bitcoin mining to AI data center development, closed a $2 billion equity round this week at a valuation exceeding $10.5 billion, according to Forbes. The valuation represents a near-doubling from its April funding round, which pegged the company at $5.5 billion.

The investor roster matters more than the headline number. New participants include funds managed by Blackstone and Jane Street, joining existing backers Nvidia and Coatue. Neither Blackstone's tactical opportunities group nor Jane Street operates venture portfolios. Their participation signals a shift in how AI compute infrastructure is being underwritten—not as speculative technology plays, but as financeable assets comparable to ports or power generation facilities.

Why it matters

The composition of Firmus's investor base reveals that AI data centers have crossed into a new asset class. When trading firms and private equity credit desks write checks, they are underwriting cash flows and collateral, not growth narratives. This marks a structural change in how capital markets view GPU infrastructure, with implications for both the durability of AI buildout and the concentration of risk between equipment suppliers and facility operators.

Sovereign compute ambitions in Asia-Pacific

Firmus is executing Project Southgate, a multi-site AI infrastructure program spanning Tasmania, Melbourne, Sydney, Canberra, and Perth. The company targets 1.6 gigawatts of AI compute capacity across five locations by 2028, partnered with CDC Data Centres. The first phase includes a 150-megawatt Melbourne facility equipped with 18,500 Nvidia GB300 GPUs, part of a broader program projected to reach A$73.3 billion in investment through 2028.

Co-chief executive Oliver Curtis stated the new capital "allows us to move on multiple fronts at once" while "fast-tracking our capacity to expand into the wider Asia-Pacific region." Beyond Australia, Firmus plans expansion into Indonesia and other regional markets, positioning itself as a provider of sovereign AI compute infrastructure powered primarily by renewable energy.

Leverage concentrates at the operator layer

Equity represents only part of Firmus's capital structure. In February, the company secured a $10 billion debt facility led by Blackstone Tactical Opportunities and Blackstone Credit & Insurance, with Coatue participating. This debt-to-equity ratio mirrors patterns across GPU cloud operators: CoreWeave, the largest U.S. player in this category, carried $25.1 billion in debt and $10.1 billion in lease obligations at the end of the first quarter.

The financing structure creates an asymmetry in how risk and reward distribute across the AI infrastructure stack. Hardware suppliers like Nvidia receive payment when equipment ships, regardless of eventual facility utilization. Operators like Firmus, by contrast, absorb years of utilization risk and refinancing exposure on borrowed capital. Equipment vendors collect certainty; facility operators bet on future tenant demand.

The broadening base of AI infrastructure investors—spanning sovereign-adjacent platforms, insurance capital, and institutional credit—suggests the buildout no longer depends on a handful of U.S. hyperscalers maintaining capex discipline. Geographic and financial diversification makes the infrastructure cycle more durable, even as leverage concentrates risk at the operator level.

Details of the funding round were first reported by Forbes contributor Jon Markman.

#ai infrastructure#data centers#firmus#blackstone#nvidia#gpu compute

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Startups

Startups· 3 min read

Snorkel AI Raises $350M to Build Data Infrastructure for Frontier Models

The Stanford spinout now valued at $3.5B supplies training data and evaluation environments to leading AI labs as model development enters a new complexity phase.

Via AI Watch · Sep 22, 2026
Startups· 2 min read

Snorkel AI hits $3.5B valuation on $350M raise

The data-as-a-service startup's revenue surged 18-fold in a year as AI labs race to secure high-quality training data.

Via AI Watch · Sep 22, 2026
Startups· 3 min read

Andreessen Horowitz Launches Tuition-Free AI Academy for 2027

The venture capital firm's one-year program targets high school graduates and early college students, with backing from OpenAI, Google, and Meta.

Via AI Watch · Sep 22, 2026