A planned expansion of Australian AI capacity

NVIDIA announced on 9 September 2026 that it is working with Australian infrastructure providers on a buildout of up to 2 gigawatts by 2027. The programme concerns sites, electricity supply and buildings intended to accommodate successive generations of NVIDIA DSX infrastructure. The figure describes planned capacity, rather than a measurement of systems already commissioned.

Participating operators include Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk. NVIDIA says the providers will run the facilities, while it supplies the computing platform, networking, software and support. This division places facility delivery and operation with the infrastructure partners.

Source: NVIDIA announces Australian AI infrastructure expansion

Connecting compute to customers

Megaport confirmed its participation in a separate announcement dated 10 September. It described the collaboration as complementary to its global network and distributed AI inference platform. The company also pointed to its acquisition of Latitude.sh, which added bare-metal accelerated computing to its network services, and to the addition of storage services this year.

Source: Megaport confirms its collaboration with NVIDIA in Australia

The engineering context

The underlying design work predates this announcement. On 16 March 2026, NVIDIA introduced the Vera Rubin DSX reference design, covering compute, Ethernet networking and storage alongside guidance for power, cooling and control systems. Its accompanying Omniverse DSX Blueprint supports digital models that simulate facility layouts, electrical arrangements and thermal behaviour before physical deployment.

Source: NVIDIA introduces the Vera Rubin DSX reference design

Procurement implications: LANGOM analysis

For procurement teams, the useful question is when a particular site can support a defined workload. A regional capacity target does not establish a delivery date for an individual project. Supplier discussions should connect the proposed server configuration with the facility schedule, cooling requirements, network access and acceptance testing.

For systems integrators, a practical next step is to request one coordinated design review covering those interfaces. Keeping the equipment specification and the commissioning plan together can expose unresolved dependencies before orders are placed.

Sources

  1. NVIDIA announces Australian AI infrastructure expansion2026-09-09
  2. Megaport confirms its collaboration with NVIDIA in Australia2026-09-10
  3. NVIDIA introduces the Vera Rubin DSX reference design2026-03-16

Planning an equipment project?

Share the application, required configuration and delivery destination with LANGOM.

Discuss your requirements

← All industry news