NVIDIA and Australian partners target up to 2GW of DSX AI factories by 2027

NVIDIA said it is working with Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk on Australian infrastructure designed for NVIDIA DSX AI factories, targeting an aggregate buildout of up to 2 gigawatts by 2027.

If delivered, the plan would expand the country’s AI-computing infrastructure across data-center capacity, power, cooling, networking and accelerated computing. The company’s announcement presents 2GW as a forward-looking target across the eight operators, not as capacity already operating.

NVIDIA did not give a partner-by-partner megawatt allocation, a facility-level commissioning schedule or arithmetic showing how the projects add up to 2GW. For comparison with a deployment framed in GPUs rather than megawatts, DataPhoenix has separately covered NVIDIA and AWS’s GPU expansion plan.

NVIDIA said the providers will operate the AI factories, while it will supply the DSX platform, accelerated computing, networking, software and ecosystem support. DSX spans facility infrastructure, compute, networking, software, simulation and reference designs, coordinating the components needed to build and run an AI facility rather than referring to a single GPU product.

Firmus is set to expand Project Southgate. Its previous Southgate plan called for up to 1.6GW of operational AI factories through 2028, including initial stages of up to 150MW in Tasmania and Melbourne. Those figures belong to the existing Southgate plan and were not identified as Firmus’s allocation within NVIDIA’s 2GW-by-2027 target.

Sharon AI is deploying up to 68,000 NVIDIA GPUs connected with Quantum InfiniBand and Spectrum-X Ethernet using DSX, NVIDIA said. The announcement did not assign Sharon AI a megawatt figure or explain how that GPU count translates into the aggregate target.

IREN is expected to combine DSX architecture with its power, land, data-center, GPU-deployment and operating capabilities, including at its Bundey campus in South Australia. IREN’s annual report lists Bundey at 800MW of planned gross capacity and describes risks involving planning, grid connection, water, permits, financing and construction. The filing does not establish that all 800MW is operating or allocated to NVIDIA’s 2027 total.

Megaport will pair NVIDIA-powered compute from its wholly owned Latitude.sh subsidiary with its software-defined network, with an emphasis on inference and lower-latency access. Megaport announced its acquisition of Latitude.sh in November 2025. ResetData’s role is to provide NVIDIA accelerated computing through an Australian AI-cloud platform for enterprise, government and research customers seeking high-performance computing and Australian data residency.

CDC will provide high-density data-center infrastructure and liquid cooling. NVIDIA cited CDC as operating more than 550MW across Australia and New Zealand, with another 800MW under construction, while CDC has separately identified itself as the data-center provider for Project Southgate. Those regional portfolio figures were not disclosed as an Australia-only or DSX-only share of NVIDIA’s target.

NEXTDC will supply power, cooling, connectivity and physical infrastructure for high-density AI factories. Its separate 3GW development pipeline and NVIDIA DGX-Ready facilities were not identified as NEXTDC’s contribution to the 2GW target. AirTrunk will expand AI-ready powered shells and direct-to-chip liquid-cooling infrastructure; its five Australian campuses in Sydney and Melbourne offer more than 755MW, but AirTrunk did not characterize that installed portfolio as a DSX allocation or incremental capacity within the target.

The announcement does not define whether its 2GW figure means energized IT load, utility-secured power, powered-shell capacity, construction underway or a broader development pipeline. It also does not separate net-new capacity from existing portfolios and previously announced projects.