Firmus Technologies Group selects VAST AI Operating System to power sovereign AI factories in Asia Pacific

Firmus Technologies Group has selected the VAST AI Operating System as the foundational data layer for its next generation of sovereign, energy-efficient AI factories across Asia Pacific. The decision reflects Firmus’ strategy to build large-scale AI infrastructure capable of supporting public-sector and regulated workloads while balancing performance, sustainability, and long-term scalability.

Founded in Australia and operating across Asia Pacific, Firmus specialises in AI infrastructure designed to maximise performance per watt. As a NVIDIA Cloud Partner, the company is building its platforms on the NVIDIA Cloud Partner reference design, aligning compute, networking, storage, and orchestration into a cohesive system. This approach is intended to support both anchor enterprise tenants and government-backed deployments as AI adoption accelerates across the region.

Firmus is already operating AI factory deployments at the scale of thousands of GPUs, with a roadmap that extends to hundreds of thousands of GPUs over time. These environments are designed for large-scale training, fine-tuning, and inference workloads, where data volumes are expected to reach petabyte scale. At this level, infrastructure efficiency is not treated as an optimisation layer but as a prerequisite for economic viability.

The company’s regional expansion includes Project Southgate in Tasmania, a flagship AI factory development powered by renewable energy. The project is positioned as a demonstration of how sovereign AI workloads can be supported under real-world energy constraints, while maintaining the performance characteristics required for advanced AI applications.

Designing AI factories as vertically integrated systems

Firmus’ infrastructure model is based on the idea that AI factories must be engineered as vertically integrated systems rather than as collections of loosely coupled components. This philosophy underpins its Model-to-Grid architecture, which integrates model behaviour, GPU performance, thermal dynamics, and grid conditions into a single optimisation framework.

“Firmus was founded on the belief that AI infrastructure must be engineered as a single, vertically integrated system,” said Daniel Kearney, Chief Technology Officer at Firmus. “Our Model-to-Grid architecture integrates model behaviour, GPU performance, thermal dynamics, and grid conditions into one optimisation framework, making our AI factories both model-aware and grid-aware, with real-time responsiveness to energy pricing and broader grid signals.”

According to Firmus, the challenge at scale is not simply deploying more GPUs, but ensuring that compute, energy, and data systems evolve in lockstep. As GPU architectures advance on increasingly compressed timelines, infrastructure must be able to absorb generational changes without introducing bottlenecks or requiring wholesale redesigns. Firmus’ AI factory model is intended to allow compute, networking, cooling, and storage layers to scale independently while continuing to operate as a unified system.

Sovereign AI is a central pillar of this strategy. Firmus’ roadmap includes in-country deployments designed to keep sensitive data under local control and to align AI infrastructure with national energy and sustainability priorities. Australia is positioned as a key anchor market due to its renewable energy resources and capacity for large-scale expansion, complementing Firmus’ established presence in Singapore and other Asia-Pacific hubs.

Data as the foundation for large-scale AI efficiency

As Firmus scales its AI factory environments, the data layer has emerged as a critical constraint on performance and efficiency. Large-scale training and inference workloads depend on sustained data throughput to keep GPUs fully utilised. Any inefficiency in data movement can cascade into higher energy consumption and reduced economic returns.

To address this, Firmus selected the VAST AI Operating System as a unified data foundation capable of operating at extreme throughput and capacity. The platform is designed to support large, disaggregated GPU clusters while sustaining efficiency as model sizes, datasets, and workload complexity increase. It also supports secure multi-tenant environments, a requirement for hosting a mix of enterprise and public-sector workloads.

“The data layer has to scale in lockstep with compute and energy,” said Kearney. “At the scale of thousands of GPUs, small inefficiencies compound quickly. We selected the VAST AI Operating System because it is architected for that reality: high throughput, disaggregated, aligned with NVIDIA Cloud Partner reference design environments, and built to sustain GPU efficiency as we expand sovereign AI capacity.”

From VAST’s perspective, the collaboration highlights how data, energy, and compute economics converge at scale. “At this scale, eliminating inefficiency isn’t a rounding error, it’s the business model,” said Jeff Denworth, Co-Founder at VAST Data. “When you’re designing AI factories to operate across tens or hundreds of thousands of GPUs, energy, data movement, and compute efficiency become inseparable.”

As Firmus continues to expand its AI infrastructure footprint across Asia Pacific, the VAST AI Operating System will underpin its data layer, supporting a transition from training to inference and beyond. The partnership reflects a broader shift in the region towards sovereign, energy-aware AI infrastructure designed for long-term operation rather than short-term experimentation.