NVIDIA is working with eight Australian data center and cloud providers to build up to 2 gigawatts of AI computing capacity by 2027, a buildout designed to meet surging demand from local startups, enterprises, and research institutions. The collaboration spans land, power, and shell capacity engineered to host multiple generations of NVIDIA DSX AI factories, bringing high-performance compute directly to Australian soil.
The participating providers - Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC, and AirTrunk - will operate the facilities. NVIDIA supplies the DSX platform, accelerated computing, networking, software, and ecosystem support. "AI factories turn energy into intelligence - the essential resource of the AI economy," said Raj Mirpuri, vice president of global AI clouds and infrastructure ecosystem at NVIDIA.
Who is building what
Firmus is expanding its Project Southgate initiative to develop AI factories serving hyperscale and AI-native demand across Australia and Asia-Pacific. "Together with NVIDIA, we are developing AI factories that will give Australian businesses, researchers and innovators access to world-class computing capacity at home," said Oliver Curtis, cofounder and co-CEO of Firmus.
Sharon AI plans to deploy up to 68,000 NVIDIA GPUs connected with NVIDIA Quantum InfiniBand and Spectrum-X Ethernet networking. James Manning, cofounder and CEO, said the company is "equipped to bring gigawatt-scale AI factories to Australian enterprises, researchers, government and AI natives." IREN is applying NVIDIA's DSX reference architecture across its development portfolio, including an 800-megawatt campus at Bundey in South Australia.
Megaport is expanding access through its subsidiary Latitude.sh, combining compute with its global software-defined network to reduce latency for inference workloads. ResetData is building infrastructure with Australian data residency for enterprises, government, and research customers. CDC has developed AI-ready facilities with direct liquid-to-chip cooling, operating more than 550 megawatts of capacity with another 800 megawatts under construction. NEXTDC and AirTrunk are both expanding liquid-cooled, high-density infrastructure designed for compute-intensive AI systems.
Software and models running on the infrastructure
The DSX platform is compatible with the CUDA ecosystem and receives software enhancements over the life of the infrastructure. NVIDIA said this approach improves utilization, extends asset life, and strengthens long-term economics for operators. The buildout also gives Australian organizations access to NVIDIA Nemotron open models for language, reasoning, speech, and multimodal applications.
Healthcare startup Heidi is combining Nemotron models with its clinical training data to build voice-driven AI tools. "NVIDIA Nemotron 3 Ultra gives us the flexibility to build and post-train models for the real complexity of healthcare," said Yu Liu, cofounder and chief technology officer. Atlassian is using NVIDIA AI to advance semantic search in Rovo, its enterprise search product. Guihong Cao, head of engineering for Rovo + AI, said the infrastructure and open model ecosystem are "foundational to how Atlassian builds and delivers AI."
Why this matters for IT, development, and government professionals
For IT and development teams inside Australian enterprises and government agencies, the 2-gigawatt target translates to locally hosted compute that sidesteps data sovereignty concerns and latency penalties from routing workloads overseas. The multiple NCPs mean competitive options rather than a single-provider lock-in. Government technology planners get a clearer timeline - 2027 - for when domestic AI factory capacity reaches hyperscale levels, which affects procurement strategies, cloud architecture decisions, and compliance planning for sensitive workloads. Developers working with CUDA and Nemotron models gain onshore infrastructure to train and fine-tune models without exporting data.
Your membership also unlocks:
