On 10 September 2026, Nvidia said it was working with Australian data-centre operators on a buildout of up to 2 gigawatts (GW) of AI capacity by 2027. If the full target is new and delivered, Australia’s total data-centre capacity could rise from about 1.6GW to more than 3.6GW—over twice today’s level.
The announcement puts Australia in the middle of the global race to build the physical infrastructure behind artificial intelligence. Nvidia named Australian cloud and data-centre companies including Firmus, CDC Data Centres, NEXTDC and AirTrunk as partners in the expansion.
This is not one giant Nvidia-owned facility. The local companies would build and operate the “AI factories,” while Nvidia would provide its DSX platform, accelerated-computing systems, networking and software. The goal is to create the dense computing capacity needed to train AI models and run the millions of daily requests sent to AI assistants, agents and industrial systems.
A buildout larger than today’s entire Australian market
The number that makes this announcement stand out is 2 gigawatts. A gigawatt is one billion watts; in this context, the figure describes the scale of electrical capacity available to computing facilities, not a promise that every site will draw its maximum continuously. Data Centres Australia estimates the country’s existing capacity at about 1.6GW, according to research cited by Reuters. Nvidia’s target is equivalent to 125% of that current base.
That comparison needs an important qualification. “Up to 2GW” is a target across several partners, not evidence that every facility has been financed, approved, connected to the grid or filled with GPUs. Large data-centre projects often arrive in phases, and announced capacity can be very different from the amount actually operating on day one.
Why Nvidia wants more Australian capacity
AI has shifted the limiting factor in computing. For many companies, the main problem is no longer finding an AI model—it is securing enough processors, electricity, cooling and network capacity to run that model at scale.
Australia offers large development sites, a sophisticated financial system, expanding renewable-energy resources and proximity to fast-growing Asia-Pacific markets. Local data centres may also appeal to governments and businesses that want sensitive information processed inside Australia rather than sent offshore.
Nvidia said the expansion would let Australian innovators develop models, applications and AI agents locally. That could matter for sectors in which Australia already has deep expertise, including mining, agriculture, healthcare, defence, climate science and financial services.
The announcement also strengthens the position of Australia’s home-grown infrastructure companies. NEXTDC is listed on the ASX, while Firmus has become one of the country’s most closely watched private AI-infrastructure developers. In February, Firmus said it had secured a US$10 billion debt package led by Blackstone and Coatue for Project Southgate, which it is developing with CDC and Nvidia.
The real bottleneck is electricity
Two gigawatts is a computing-capacity figure, but it is also a warning about the grid. An AI data centre can require the electricity of a major industrial complex and must receive that power continuously. New transmission, substations, generation and storage may take longer to build than the server halls themselves.
Water use is another concern where facilities rely on evaporative cooling. Communities will reasonably ask who pays for grid upgrades, how much water is consumed, whether new renewable generation is genuinely additional and what happens during heatwaves or periods of tight electricity supply.
Nvidia says the expansion can support additional power-generation projects, but that outcome is not automatic. It will depend on the contracts signed by individual operators and on planning decisions by governments and network companies.
The pressure is not unique to Australia. Google’s planned European AI expansion, covered recently by FutureTechDose, is linked to electricity demand comparable with a large share of a nuclear power station.
Read: Google’s €13 Billion AI Bet Comes With Half a Nuclear Plant
What Australians could actually gain
The clearest near-term benefits would be construction work, electrical and mechanical engineering, network infrastructure and long-term technical roles operating the facilities. More local compute could also reduce latency and make it easier for Australian organisations to keep regulated data within the country.
The larger economic prize is harder to guarantee. Owning server buildings does not automatically create globally competitive Australian AI companies. To capture more value, Australia would also need researchers, software developers, affordable access to computing, commercial customers and rules that encourage innovation while protecting data and communities.
Australia can become either a major AI producer or simply a well-powered location where overseas companies place machines. The distinction will depend on who receives access to the infrastructure and what is built on top of it.
What happens next
The 2027 target makes the next 15 months critical. The strongest evidence of progress will not be another headline—it will be named sites, approved power connections, construction milestones, binding customer contracts and operational Nvidia systems.
For now, the confirmed development is that Nvidia has publicly named its Australian partners and set an unusually large capacity target. The opportunity is real, but the final scale, timing and environmental footprint remain to be proven.
Australia’s recent quantum-battery breakthrough showed that local science can attract international attention. Nvidia’s new push asks a bigger question for the economy: can Australia build not only clever inventions, but also the immense infrastructure required to turn AI into an industry?
Sources
- Reuters: Nvidia plans major expansion of data-centre capacity in Australia, 10 September 2026.
- Reuters: Firmus lands US$10 billion debt package, 9 February 2026.
- Reuters: Nvidia-backed Firmus signs OpenAI capacity deal, 8 September 2026.
This article is for general information and does not constitute financial or investment advice.
Leave a Reply