NVIDIA Partners on Australian AI Capacity Buildout Targeting 2 Gigawatts by 2027

NVIDIA has joined Australian cloud and infrastructure providers in a plan to build capacity for AI factories—large facilities designed to run AI computing—with a target of up to 2 gigawatts by 2027. The effort would give Australian organizations more local access to accelerated computing and NVIDIA’s Nemotron open models, while also serving regional and global AI-compute demand.

The announcement arrives as NVIDIA says growing AI adoption is increasing demand for high-performance computing. Rather than build and run the sites itself, NVIDIA is collaborating with Australian NVIDIA Cloud Partners and infrastructure providers that will operate the facilities. NVIDIA’s role is to supply its DSX platform, accelerated computing, networking, software and ecosystem support.

A facility plan built around several layers

DSX is NVIDIA’s full-stack AI-factory platform, spanning facility infrastructure, computing, networking, software and reference designs. NVIDIA says the platform is compatible with its CUDA software ecosystem and can be enhanced through software over the life of the infrastructure, an approach intended to let operators support multiple generations of NVIDIA compute.

Up to 2Australian buildout target

NVIDIA described the figure as a planned buildout, not current operating capacity.

Up to 68,000Sharon AI deployment

Sharon AI says its deployment will connect the GPUs through NVIDIA Quantum InfiniBand and Spectrum-X Ethernet networking.

Eight named participants are Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk. Their projects illustrate that the initiative is not one uniform construction program: some involve data-center facilities and cooling, while others center on GPU deployments, connectivity or AI-cloud services.

Three disclosed pieces of the buildout

• Sharon AI is deploying up to 68,000 NVIDIA GPUs with NVIDIA Quantum InfiniBand and Spectrum-X Ethernet networking.
• IREN says it is applying the DSX architecture across its development portfolio, including its 800-megawatt Bundey campus in South Australia.
• CDC says it operates more than 550 megawatts across Australia and New Zealand, has another 800 megawatts under construction, and uses direct liquid-to-chip cooling.

Local models are the intended payoff

NVIDIA says the added capacity is intended for startups, universities, researchers, enterprises, developers and AI-native companies building models, applications and agents for Australian industries and communities. The company identifies Heidi and Atlassian as organizations using Nemotron open models, datasets and tools for healthcare and enterprise-search work, respectively.

The next test is execution. The announced 2-gigawatt figure is explicitly a target for 2027, and the program depends on separate operators bringing sites and infrastructure online while NVIDIA supplies the technical stack. If it proceeds as planned, the arrangement would expand local computing access without limiting the facilities to domestic demand.