NVIDIA Plans Up to 2 GW of AI Infrastructure in Australia With Eight Operators by 2027
Редакция ЦОД:Главное
NVIDIA has announced a framework with eight Australian cloud and data-center infrastructure operators to prepare up to 2 GW of AI infrastructure by 2027. The figure is a maximum programme target, not operational capacity or a confirmed construction pipeline.
- The announced programme targets up to 2 GW of Australian AI infrastructure by 2027; it is not a statement of operating or fully committed construction capacity.
- Eight operators are expected to expand land, power and building resources for NVIDIA DSX AI-factory deployments.
- Power generation availability and grid interconnection capacity are identified as the principal constraints on delivery.
- IREN, CDC and Sharon AI have separately described capacity plans, but these should not be treated as fully allocated to the NVIDIA programme.
- NVIDIA’s claimed up-to-40% gain in GPU capacity within a fixed facility power limit is a design estimate, not a reported project outcome.
NVIDIA has announced collaborations with Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC and AirTrunk to prepare up to 2 GW of AI infrastructure in Australia by 2027.
The arrangement is a framework for expanding the operators’ available land, power supply and buildings to host multiple generations of NVIDIA DSX AI factories. NVIDIA will provide the DSX platform, accelerated computing, networking, software and ecosystem support. The 2 GW figure is the upper limit of the overall programme; it does not represent commissioned capacity, nor does it confirm that the full volume has been contracted for construction.
The participating operators bring distinct existing or separately announced development pipelines. IREN is developing the Bundey campus in South Australia with planned total-project capacity of 800 MW, although the available information does not establish that all of this capacity is allocated to the NVIDIA collaboration. CDC operates more than 550 MW of data-center capacity across Australia and New Zealand and has a further 800 MW under construction. Sharon AI had previously announced a six-year NVIDIA collaboration covering 72 MW of new data-center capacity and up to 40,000 Grace Blackwell GB300 GPUs.
Sharon AI has also indicated 132 MW of AI-factory capacity following that collaboration, with 102 MW stated as contracted with customers. Its broader plans include up to 68,000 NVIDIA GPUs, a figure that covers multiple deployment plans rather than solely the additional 40,000 GPUs in the previously announced agreement.
For operators, the programme underscores that AI-ready capacity is increasingly a site-and-power delivery challenge. NVIDIA says DSX design approaches can increase the GPU capacity accommodated within a fixed facility power envelope by up to 40%, but this is a company modelling estimate and will depend on site configuration and workload characteristics.
The programme could accelerate standardisation of AI-ready data centers around an integrated approach to compute, liquid cooling, power delivery and demand management. In the Australian market, this could place greater value on sites with demonstrable energy resources and grid interconnection, while leaving the programme’s ultimate delivery dependent on generation availability and connection capacity.
The initiative shifts the AI-infrastructure discussion beyond GPU supply toward coordinated delivery of sites, power, cooling, networking and load management. Its execution will depend primarily on available generation and grid interconnection capacity, rather than compute-equipment demand alone.
- — / 11 сентября 2026 г. · ЯпонскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.