Firmus, an Australian company that began as a Bitcoin mining operation, has transformed itself into a $10.5 billion AI infrastructure developer after securing $2 billion in fresh equity funding. The funding round, which nearly doubled the company's valuation from $5.5 billion in April, reveals a fundamental shift in how capital markets evaluate AI computing assets: not as speculative tech ventures, but as long-term infrastructure comparable to power plants and toll roads.

Why Are Non-Traditional Investors Suddenly Backing AI Data Centers?

What makes Firmus's latest funding round remarkable isn't just the valuation jump, but the identity of the investors writing the checks. The round included existing backers Nvidia and Coatue, but also brought in funds managed by Blackstone and Jane Street, neither of which operates a traditional venture capital portfolio. Jane Street is a trading firm; Blackstone Tactical Opportunities is a private credit desk designed to acquire complex assets other investors avoid. Their participation signals a critical market shift: AI factories are being underwritten the same way infrastructure assets like transmission lines and toll roads are evaluated, based on cash flows, contracts, and collateral rather than growth stories.

This represents a broadening of the investor base beyond venture capital. Over the past year, Firmus has raised more than $3 billion in equity financing, with the company's valuation roughly doubling in just four months between April and August. Beyond equity, the company closed a $10 billion debt facility in February, led by Blackstone Tactical Opportunities and Blackstone Credit and Insurance, with Coatue also participating. That scale of debt financing, underwritten by insurance-backed credit platforms, demonstrates that AI infrastructure operators are now being treated as financeable asset classes comparable to traditional utilities.

What Is Firmus Building, and Where?

Firmus began its corporate life mining Bitcoin but has completely repurposed itself around AI infrastructure development. The company is constructing data centers built on Nvidia's DSX reference architecture and its own HyperCube hardware platform. Its flagship initiative, Project Southgate, started in Tasmania and is now expanding across multiple Australian cities: Melbourne, Sydney, Canberra, and Perth.

The scale of this buildout is substantial. Project Southgate aims to deploy 1.6 gigawatts of AI computing power across five sites by 2028 in partnership with CDC Data Centres. The initial phase, centered in Melbourne, carries an investment of AU$4.5 billion and is equipped with 18,500 Nvidia GB300 GPUs. The total investment for the entire Southgate project is expected to reach AU$73.3 billion by 2028. After Australia, Firmus plans to expand into Indonesia and the broader Asia-Pacific region, positioning itself to offer sovereign AI compute capacity powered largely by renewables without requiring customers to wait in line behind American hyperscaler demand.

How Does Risk Distribution Work in AI Infrastructure?

A critical insight emerges from examining who bears the financial risk in this boom: equipment suppliers and operators occupy fundamentally different positions. Nvidia and other hardware vendors collect revenue immediately upon GPU shipment, regardless of whether the data centers actually fill their capacity with paying customers. Firmus and other operators, by contrast, must absorb years of utilization risk and refinancing risk on borrowed money.

This risk asymmetry extends across the industry. CoreWeave, the largest American GPU cloud operator, ended the first quarter with $25.1 billion in debt and another $10.1 billion in lease obligations. The operator layer of the AI buildout runs on leverage almost everywhere, meaning companies like Firmus must borrow now, build for years, and count on tenants filling capacity that does not yet exist. Firmus pairs its borrowing with long-dated government-adjacent tenancy through CDC, which provides some stability, but the fundamental structure remains the same: operators absorb leverage while suppliers collect certainty.

Steps to Understanding AI Infrastructure as an Asset Class

• Recognize the Investor Shift: When trading firms and private credit desks invest in AI data centers, they are evaluating them as infrastructure assets with predictable cash flows, not as high-growth tech companies with speculative upside.
• Understand the Financing Structure: AI infrastructure operators typically combine large equity raises with massive debt facilities, mirroring the financing model used for toll roads, power plants, and transmission lines rather than traditional venture-backed startups.
• Identify the Risk Distribution: Equipment suppliers like Nvidia get paid upfront when hardware ships, while operators like Firmus absorb multi-year utilization risk and refinancing risk on borrowed capital, creating asymmetric risk exposure across the supply chain.
• Track Geographic Expansion: Sovereign-adjacent platforms in Australia, Indonesia, and the broader Asia-Pacific are now funding compute demand independently of American hyperscaler decisions, widening the geographic and financial base for AI infrastructure investment.

The broader implication is that the bear case on AI capital expenditure, which has long rested on the idea that a handful of American hyperscalers carry the entire buildout and will eventually cut spending, is weakening. Every new buyer class that shows up, whether sovereign-adjacent platforms, private credit desks, insurance capital, or trading firms, funds compute demand that exists whether or not any single hyperscaler trims its budget. The demand base for AI infrastructure is widening geographically and financially at the same time, and that breadth is what makes the buildout durable.

For Bitcoin miners watching this transformation, Firmus's pivot offers a cautionary and instructive tale. The company recognized that Bitcoin mining, with its razor-thin margins and commodity-like competition, could not sustain the capital intensity required to remain competitive. By transitioning to AI infrastructure, Firmus tapped into a market where capital is abundant, investors are willing to underwrite leverage, and the asset class itself is being treated as essential infrastructure rather than speculative technology. That strategic shift, more than any single funding round, explains why a former Bitcoin miner is now valued at $10.5 billion.