What Firmus Technologies does

Firmus Technologies is an AI infrastructure company originally founded in Australia and now headquartered in Singapore. The company designs, builds, and operates modular, vertically integrated "AI Factories" — purpose-built data center facilities engineered for high-density artificial intelligence training and inference workloads. Its customer base comprises hyperscale cloud providers, sovereign AI initiatives, and government entities across the Asia-Pacific region and the Middle East.

The technical stack is composed of four interlocking components: a proprietary direct-liquid cooling platform branded Hypercube, an orchestration and management software layer called AI FactoryOS, NVIDIA GPU hardware deployed at scale, and a workload optimization and market-facing software suite called Model-to-Grid. This combination targets the thermodynamic, density, and energy constraints of modern AI workloads, positioning Firmus as a vertically integrated alternative to traditional colocation providers and pure-play GPU cloud operators. The company is the lead commercializer of NVIDIA Blackwell systems in Australia and is building a multi-site Asia-Pacific footprint.

Revenue is generated through two distinct mechanisms. The first is long-dated, multi-year infrastructure contracts that anchor sovereign and hyperscale customers, exemplified by Project Southgate with the Tasmanian Government and a $1.7B sovereign contract with the UAE. The second is an on-demand AI cloud service branded Sustainable Metal Cloud, which offers pay-as-you-go GPU access alongside emerging token-metered inference pricing. The business is highly capital-intensive: Firmus has secured a US$10B debt facility from Blackstone, ~US$505M of equity from Coatue at a US$5.5B valuation, and A$100M from Maas Group, and is preparing an ASX listing targeting an approximately A$2B primary raise to fund continued facility build-out.

Name

Firmus Technologies

Legal name

Firmus Technologies Pty Ltd

Website

https://firmus.co

Company type

Private

Founded year

2019

Operating status

Operating

Headcount range

101–250 employees

Short description

Firmus Technologies designs, builds, and operates modular AI Factory data centers combining proprietary liquid cooling, NVIDIA GPUs, and orchestration software to serve hyperscale, sovereign, and government customers across the Asia-Pacific region.

Ownership category

akta.pro rank

Product category

AI Cloud Infrastructure

NAICS

Computer and Peripheral Equipment Manufacturing (33411)

SIC

Electronic Computers (3571)

akta.pro primary industry

AI Server Systems & HGX/Accelerator Platforms (HDAAAAAB)

akta.pro secondary industries

AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI), AI Compute Cloud & GPU-as-a-Service (HDAAAAAK), End-to-End Enterprise AI Platforms (MLOps & Model Lifecycle Management) (HDAEANAA)

Keywords

• AI cloud services
• GPU cloud infrastructure
• Liquid cooling data centers
• AI factory design
• Sovereign AI compute

Headquarters

HQ city

Saint Leonards

HQ country

Australia

HQ region

Oceania

Offices10 records

Markets served

GTM type

B2B

Offering type

Services

Cost components

Technology or R&D, Infrastructure, Supply Chain, Personnel, Operations, Marketing or Sales

Revenue model

1. AI Cloud Services (Sustainable Metal Cloud / SMC): GPU-powered AI cloud services offered as on-demand compute, bare metal clusters, AI storage, and pre-built cloud applications. Customers reserve or consume resources for AI training, inference, and HPC workloads. Reservation and on-demand models available.
2. AI Cloud Platform-as-a-Service (via Rafay integration): Self-service AI development environment including Jupyter Notebooks, ML Workbenches, NVIDIA NIM, serverless inference, and AI Studio for model hosting, streaming fine-tuning and inferencing.
3. Token-based AI Compute Metering: Adoption of Rafay's Token Factory platform enables token-metered access to AI models running on GPU infrastructure, positioning Firmus to compete beyond commodity hourly pricing.
4. Long-term Hyperscale AI Factory Contracts: Multi-year infrastructure agreements with global hyperscale customers at Project Southgate campuses (e.g., multi-billion dollar Melbourne data center with ~18,400 Nvidia GB300 chips for unnamed hyperscale customer).
5. AI Factory Infrastructure Build-out (Capital Project): Capital deployment model for sovereign AI factory campuses, funded by equity and debt, with operational revenue flowing once facilities go live (e.g., Project Southgate Stage 1a with $4.5 billion investment in Tasmania; Southgate Melbourne with initial 18,500 NVIDIA GB300 GPUs live Q1 2026).

Pricing tiers

Go-to-market motion4 records

Distribution channels7 records

Marketing channels8 records

Core offering

Firmus Technologies designs, builds, and operates vertically integrated "AI Factories"—modular GPU-compute facilities powered by proprietary HyperCube liquid cooling—and delivers GPU-as-a-Service through its AI Cloud platform (branded Sustainable Metal Cloud / SMC). It combines the physical AI Factory infrastructure with the AI FactoryOS orchestration software and Model-to-Grid energy-aware workload management to deliver token-metered AI compute to hyperscalers, sovereign governments, and enterprise customers across Australia and Singapore.

Product overview

Firmus Technologies operates a vertically integrated, platform-plus-modules AI infrastructure offering. The two flagship pillars are Firmus AI Cloud (the GPU-powered neocloud delivered through sub-products Cloud Compute, Bare Metal, AI Storage, Cloud Services and Cloud Applications — supported by the Sustainable Metal Cloud/SMC sub-brand in Singapore) and the Firmus AI Factory (the physical modular infrastructure built on the proprietary HyperCube unit of abstraction, deployed at scale through Project Southgate in Australia and via the Loyang and Media Hub sites in Singapore). Both are unified by AI FactoryOS™, the proprietary orchestration and telemetry layer, and by the Model-to-Grid software approach that integrates the AI Factory stack with NVIDIA DSX Blueprint for grid-aware compute orchestration. Connectivity, supply chain, subsea fibre (Bernacchi-1) and R&D (Research Lab) round out the offering, with NVIDIA NIM, Rafay, Megaport, Netris NAAM, VAST Data and WEKA serving as third-party technical integrations.

Differentiator

Problem solved

Functional benefit

Brands

• Sustainable Metal Cloud (SMC): AI-native cloud platform from Firmus (later rebranded/expanded as Firmus AI Cloud), providing GPU cloud services for sovereign AI workloads.
• Project Southgate
• AI FactoryOS
• Firmus HyperCube
• Model-to-Grid

Products and services

• Firmus AI Cloud Token-metered, on-demand GPU compute cloud platform delivering AI training and inference capacity to enterprise and AI-native customers across Australia, Singapore, and global markets. Combines Cloud Compute, Bare Metal, AI Storage, and Cloud Applications modules.
• Firmus AI Factory Modular, vertically integrated GPU compute facility combining proprietary HyperCube liquid cooling, AI FactoryOS orchestration, and grid-aware power systems. Designed, built, and operated by Firmus for hyperscale, sovereign, and enterprise customers.
• Project Southgate Flagship hyperscale AI Factory deployment programme built around NVIDIA DSX Vera Rubin reference architecture, serving as the anchor project for Firmus's large-scale built-to-suit pipeline.
• AI FactoryOS Proprietary orchestration and telemetry operating system that runs across every Firmus AI Factory, providing governance, workload automation, system-wide visibility, and the Model-to-Grid energy-aware scheduling layer.
• Sustainable Metal Cloud (SMC) Customer-facing brand of the Firmus AI Cloud platform providing sustainable, liquid-cooled GPU compute services to AI developers, researchers, and enterprises.
• Firmus Cloud Compute On-demand, token-metered GPU compute service module of the Firmus AI Cloud for AI training, fine-tuning, and inference workloads.
• Firmus Bare Metal Reserved bare-metal GPU server offering within the Firmus AI Cloud for customers that require dedicated, single-tenant hardware with predictable performance.
• Firmus AI Storage High-performance data storage service module of the Firmus AI Cloud, designed to feed large-scale AI training and inference workloads, integrated with the WEKA Augmented Memory Grid.
• Firmus Cloud Services Managed cloud services layer of the Firmus AI Cloud, providing operational support, deployment, and managed-service capabilities around the compute and storage modules.
• Firmus Cloud Applications Cloud-based AI applications layer of the Firmus AI Cloud, offering ready-to-use AI software and tools on top of the underlying compute and storage services.
• Model-to-Grid Patented energy-aware workload management approach that schedules AI compute jobs in alignment with real-time grid conditions, renewable energy availability, and workload priority, underpinning AI FactoryOS.
• Bernacchi-1 Submarine Cable Submarine cable connectivity asset linking Firmus's AI Factory data centers to provide high-bandwidth, low-latency interconnect between sites.

Quantifiable outcome

• 45% better FLOP per utility picoJoule vs traditional data centers
• +9 more outcomes

Named customers5 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration11 records

AI capability11 records

Feature6 records

Partnerships

Ten partnerships are on record, tiered major - strategic infrastructure partner, flagship, major - board appointments for ipo readiness, flagship - data center co-development partner, major - long-term manufacturing partner, flagship - flagship singapore partnership and customer, major, major - global connectivity partner and major - singapore government partnership.

• SUBCOmajor - strategic infrastructure partnerStrategic or Co-development Partner · 2 June 2026Partnership to build Bernacchi-1, a submarine fibre optic cable connecting Tasmania to mainland Australia via SUBCO's SMAP cable system. Firmus underwrites the branch and provides capital. First new Tasmania-mainland cable in 20+ years; 60+ Tbps capacity; operational Q2 2027. Founder Bevan Slattery publicly endorsed.
• NVIDIAflagshipStrategic or Co-development Partner · 7 April 2026NVIDIA is a multi-round equity investor (participating in $505M round April 2026, $330M September 2025, $327M November 2025 round) and technology partner. Firmus's AI Factory infrastructure is built in alignment with NVIDIA Vera Rubin DSX reference design and uses NVIDIA GPUs (H200, H100, GB300, Vera Rubin). Co-developed grid-integrated AI Factory software with NVIDIA DSX Blueprint. Firmus is an NVIDIA Cloud Partner (NCP), leveraging NVIDIA Spectrum-X Ethernet, NVIDIA Nemotron open models, and NVIDIA NIM. Marc Hamilton (VP Solutions Architecture and Engineering, NVIDIA) is a public endorser.
• Lee Hatton, Christine Bartlett, Julie Shuttleworth AMmajor - board appointments for ipo readinessStrategic or Co-development Partner · 24 March 2026Three new Non-Executive Directors appointed to Firmus Grid Limited board: Lee Hatton (Chair of Audit and Risk Committee, ex-Xero board, NAB, Suncorp, Afterpay, Block Inc); Christine Bartlett (Chair of Remuneration and Nomination Committee, ex-IBM, NAB, Jones Lang LaSalle, current Chair NSW Ports and CEDA); Julie Shuttleworth AM (ex-Deputy CEO Fortescue, CEO Fortescue Future Industries). Mike Ferraro also appointed 21 Nov 2025 (ex-CEO Alumina Limited, Chief Legal Counsel BHP Billiton).
• CDC Data Centresflagship - data center co-development partnerStrategic or Co-development Partner · 9 February 2026Co-development partner for Project Southgate to develop AI training and inference data centers across Australia. CDC and Firmus agreement targets up to 1.6GW of capacity by 2028 powered by Nvidia chips. CDC logo displayed on Firmus website under "Industry partners".
• Benmaxmajor - long-term manufacturing partnerImplementation/ SI/ Consulting Partner · 5 February 2026Five-year partnership for mechanical systems manufacturing. Firmus has worked closely with Benmax to ramp production capacity for AI Factory components including cooling, power, and integrated system modules. Manufacturing supports hundreds of long-term jobs.
• AI Singapore (AISG)flagship - flagship singapore partnership and customerStrategic or Co-development Partner · 1 December 2025MoU partnership for SEA-LION (Southeast Asian Languages in One Network) - first family of open-source LLMs for Southeast Asian languages. 32 nodes/256 H200 GPUs deployed. Partnership covers Singapore-based H200 GPU access, global hosting of SEA-LION, and joint AI benchmarking research.
• WEKAmajorTechnology or Integration · 14 November 2025Firmus deployed WEKA's Augmented Memory Grid memory extension technology combined with NVIDIA STX/Vera Rubin NVL72 architecture. Achieved 6.5x more token output per GPU without additional energy consumption. WEKA is also featured as Firmus's S3-compatible storage tier on the AI Storage product page.
• Megaport (ASX:MP1)major - global connectivity partnerChannel Partner/ Reseller/ Distributor · 1 October 2025Key connectivity partner for Project Southgate. Provides one-click national and global connectivity between Southgate's AI Factory availability zones and customer environments via Megaport's presence in 1,000+ data centers across 180+ cities. Usage-based, elastic bandwidth model. ISO 27001 certified.
• Rafay SystemsmajorTechnology or Integration · 16 September 2025Firmus selected Rafay to integrate Platform-as-a-Service capabilities into the Firmus AI Cloud. Enables GPU-powered compute (bare metal, VMs, SLURM, Kubernetes), self-service tools (Jupyter, ML Workbenches, NVIDIA NIM), serverless inference, and AI Studio. Further extended via Rafay's Token Factory for token-metered access to AI models.
• HTX (Home Team Science and Technology Agency, Singapore)major - singapore government partnershipStrategic or Co-development Partner · 1 May 2025Memorandum of Understanding to collaborate on advanced AI infrastructure design and research supporting HTX's mission to enhance Singapore's public safety systems while meeting energy efficiency goals. Joint research into liquid-cooled AI infrastructure and design principles for next-generation AI factories.

Scale indicators12 records

Recent moves7 records

Expansion highlights7 records

Direct peers

• CoreWeave: US-based neocloud built on NVIDIA GPUs (H100/H200/GB200) with vertically integrated data centers and purpose-built AI cloud software; the closest comparable to Firmus in business model, customer base (hyperscale and AI labs), and reliance on NVIDIA's roadmap.
• Crusoe Energy: Renewable-powered AI compute provider (initially flared-gas, now pivoting to renewables) that builds and operates modular AI data centers with a sustainability-led value proposition; directly comparable to Firmus's hydro-powered, liquid-cooled, ESG-positioned AI Factory model.
• Lambda: GPU cloud provider specializing in NVIDIA-based instances for AI training and inference, with both on-demand cloud and dedicated clusters; competes with Firmus AI Cloud for the same AI developer and enterprise ML workload.
• Together AI: AI-focused cloud platform offering NVIDIA GPU clusters for training and inference, with proprietary inference stack and token-based pricing; directly comparable to Firmus's Sustainable Metal Cloud in customer profile and GPU-as-a-service motion.

Emerging players

• Vultr: Global cloud provider offering GPU instances alongside bare metal; smaller and more generalist than the hyperscalers but increasingly focused on AI workloads, putting it in partial overlap with Firmus's reserved bare-metal and on-demand GPU cloud.
• Ori Industries: Distributed cloud infrastructure provider targeting AI and high-performance workloads at the edge with energy-efficient designs; thematically comparable to Firmus's regional, sustainability-driven AI factory thesis.

Broad incumbents

• Oracle Cloud Infrastructure (OCI): Hyperscale cloud with a large NVIDIA GPU footprint (H100/GB200) and explicit focus on sovereign and AI workloads; a broad incumbent offering overlapping capability to Firmus's AI Cloud and Project Southgate hyperscale contracts.
• Amazon Web Services (AWS): Largest cloud provider and the benchmark Firmus is measured against in SemiAnalysis ClusterMAX™; provides NVIDIA GPU instances at scale and has committed A$20B to Australian data centers, directly contesting Firmus's sovereign AI positioning.
• Google Cloud Platform: Hyperscale cloud offering both NVIDIA GPUs and Google's custom TPU silicon; competes with Firmus for AI training and inference workloads, particularly in APAC, and is explicitly named as a peer Firmus outranks in ClusterMAX™.

Regional players

• nLighten: European edge data center operator expanding into GPU-as-a-service and AI-ready capacity with sustainability positioning; thematically comparable to Firmus's regional, low-latency AI factory thesis but focused on Europe rather than APAC.

Market position

Strengths4 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

• LinkedInlinkedin.com/company/firmustechnologies
• Xx.com/firmuscloud

Compliance6 records

Revenue estimate

Valuation estimate

Number of profiles

Profiles14 records

Subsidiaries3 records

Funding overview

Funding rounds6 records

Investors11 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A1 record

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Firmus Technologies