Firmus builds liquid-cooled AI data centers engineered for maximum token-generation efficiency and profitability. The tech stack reveals deep operational maturity: Kubernetes + Slurm for workload orchestration, PyTorch + Megatron-LM for large-scale model training, and a full observability pipeline (ClickHouse, dbt, Metabase, Tableau). Senior-heavy engineering hiring and active projects spanning hardware failure prediction, Kubernetes services, and network automation signal a company scaling infrastructure layers, not just selling capacity.
Firmus Technologies designs and operates AI Factories — purpose-built data centers that maximize token-generation throughput and profitability through advanced thermal management and holistic energy optimization. The company innovates across the full stack: power grid integration, cooling systems, GPU and network telemetry, and operational automation. Based in Tasmania with hiring across Australia, Singapore, and the United States, Firmus serves organizations building or scaling large-scale AI compute, from training to inference workloads. The business is operationally mature, with governance and risk frameworks aligned to pre-IPO standards.
Firmus builds liquid-cooled AI data centers optimized for token-generation efficiency. Active projects include sustainable factory development, hardware failure prediction, a managed Kubernetes service, and network-as-code automation across their AI Factory infrastructure.
Engineering stack includes PyTorch, Megatron-LM, Slurm, Kubernetes, and NVIDIA H200 GPUs. Data and observability layers run on ClickHouse, dbt, Delta Lake, and Tableau. Security tooling spans Vault, Falco, Kyverno, and Snyk.
Other companies in the same industry, closest in size