Artificial intelligence is turning into a heavy industry. NVIDIA has announced a collaboration with a group of Australian data center and digital infrastructure operators to build up to 2 gigawatts of capacity dedicated to AI computing systems by 2027. Partners include Firmus, Sharon AI, IREN, Megaport, ResetData, CDC, NEXTDC, and AirTrunk.
The figure speaks louder than many presentations about how dramatically the scale of the industry has shifted. Two gigawatts represent an electricity demand that cannot be treated as a technical detail. It requires land, grid interconnections, cooling systems, fiber, transformers, power purchase agreements, and permitting timelines compatible with investments measured in the billions.
AI factories become national infrastructure
NVIDIA describes these sites as AI factories: full-stack infrastructures built around the DSX platform and designed to host multiple generations of accelerated systems. The stated goal is to provide capacity to Australian startups, universities, researchers, and enterprises, preventing access to advanced computing from remaining concentrated exclusively in the United States or a handful of major global hubs.
The geopolitical dimension is unmistakable. Having computational capacity on domestic soil means cutting latency, tightening control over data, and fostering local expertise. Countries that just a few years ago were primarily debating connectivity are now planning full-fledged AI industrial districts.
The bottleneck is increasingly physical
The project's success will also depend on Australia's ability to power it without straining grids and driving up prices. The next phase of AI will not be decided solely by the quality of the models, but by the availability of electricity, capital, and infrastructure. Without megawatts, the world's most advanced software simply grinds to a halt.
