NVIDIA and Australian Partners Target 2GW AI Infrastructure Buildout by 2027

Dedicated Servers and GPU

NVIDIA has announced a partnership with eight Australian cloud and data center companies targeting up to 2 gigawatts of AI infrastructure by 2027, a buildout that would substantially increase the country’s capacity for training and running artificial intelligence models.

The company named Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk as participants. The providers will operate the facilities, while NVIDIA will supply its DSX platform, accelerated computing systems, networking, software and broader technology ecosystem.

The plan was announced on September 9. It covers an expansion of land, power and powered-shell capacity designed to support multiple generations of NVIDIA DSX infrastructure rather than a single centrally managed data center project.

Announced capacity target

Up to 2GW by 2027

The figure is a forward-looking target across eight infrastructure partners. NVIDIA did not disclose a total investment value, individual capacity allocations or a commissioning schedule for each project.

Eight operators, one infrastructure platform

According to the official NVIDIA announcement, the DSX platform spans facility infrastructure, computing, networking, software and reference designs. The common architecture is intended to let operators support successive generations of NVIDIA hardware without treating each deployment as an isolated system.

The partners enter the initiative with different roles. Firmus is expanding its Project Southgate program, while IREN plans to apply the DSX architecture to its development portfolio, including its 800MW Bundey campus in South Australia. Sharon AI said it is deploying up to 68,000 NVIDIA GPUs connected through Quantum InfiniBand and Spectrum-X Ethernet.

CDC, NEXTDC and AirTrunk will provide the physical data center layer, including high-density power and liquid-cooling infrastructure. ResetData is targeting enterprise, government and research workloads requiring Australian data residency. Megaport, through its Latitude.sh subsidiary and global network, is focused on distributed inference capacity and connectivity.

Power becomes the critical dependency

The 2GW figure represents planned capacity rather than infrastructure already commissioned. NVIDIA’s announcement does not divide that target among the eight companies or identify how much of it has received planning approval, secured grid connections or reached construction.

That distinction matters because power availability, transmission capacity and construction schedules increasingly determine whether announced AI data center projects can enter service. NVIDIA said the expansion would support additional power-generation projects, effectively acknowledging that compute deployment and energy development will need to advance together.

Reuters reported that Australia’s existing data center capacity is approximately 1.6GW, citing Data Centres Australia and research by DC Byte. If the full additional target is delivered, the program would more than double the current national total.

The expansion also arrives as Australian policymakers face calls for closer scrutiny of data centers’ electricity and water requirements. The announced 2027 deadline leaves a relatively short window for operators to secure sites, utility capacity, cooling equipment, network connectivity and the specialized construction resources required for high-density AI deployments.

What happens next

The next measurable developments will be project-level disclosures from the participating operators: confirmed sites, grid agreements, construction starts, capacity allocations and commissioning dates. Those details will determine how much of the 2GW headline translates into operating infrastructure by 2027.

For the infrastructure industry, the announcement shows NVIDIA extending its role beyond supplying accelerators. Through DSX, it is promoting a repeatable facility-to-software architecture while relying on local operators to secure and run the physical assets.

The scale is therefore significant, but it remains a target. Delivery will depend on whether the eight partners can convert their existing development portfolios into powered, cooled and connected capacity within the stated timeframe.