According to a recent LinkedIn post from Firmus Technologies, the company’s board and executive leadership recently toured one of its newest “AI Factory” environments, which is described as concentrating more than US$1 billion of next‑generation AI compute capacity. The site is said to be built around NVIDIA GB300 systems and a HyperCube installation that is entering its final stages, indicating a large-scale, GPU-centric infrastructure build-out.
Claim 55% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
The post outlines a deployment consisting of 256 racks across four data halls, with 64 racks per hall and two HyperCube modules per hall, each configured with 2 × 16 rows of NVL72 GB300 racks. In aggregate, the facility is expected to host more than 18,000 NVIDIA GB300 GPUs within a single, optimized environment, suggesting Firmus is positioning itself as a high-density AI infrastructure provider.
As shared in the post, while the Melbourne deployment progresses, Firmus engineers are working on a GB300 development cluster in Canberra to tune the rack-scale platform for the company’s AI FactoryOS and Model-to-Grid infrastructure stack. This dual-track approach—simultaneous build-out and platform tuning—may shorten time-to-market for commercial AI workloads and enhance the company’s ability to monetize capacity once fully online.
The emphasis on “Project Southgate” as industrial-scale AI infrastructure implies ambitions to compete in the rapidly expanding market for AI data centers and cloud-like compute services. For investors, the described capital intensity and reliance on NVIDIA’s latest GPUs point to a strategy targeting high-value AI training and inference workloads, though it may also entail significant ongoing funding requirements and execution risk during the ramp-up phase.