Australia emerges as a potential AI compute hub, leveraging abundant and cheap renewable energy to offset limited existing infrastructure. With only 1.4GW of current AI compute capacity, the nation ranks mid-way globally, behind leaders like northern Virginia. However, major developments are underway. Anthropic has committed to a A$30bn ($21bn) project near Brisbane, while New South Wales advanced discussions for projects seeking 13GW — exceeding the state's current daily demand. Grid operators project data centres could consume 12 per cent of national supply by 2050, up from 2 per cent currently.

The economic proposition relies on Australia's exceptionally low renewable energy costs, with solar priced at $40 per MWh, competitive with Saudi Arabia and 35 per cent cheaper than the US. This cost advantage could help spread the financial burden of the clean energy transition. Despite planning requirements of $100bn in investment by 2030 according to Commonwealth Bank of Australia, the continent offers almost limitless space for solar and wind generation.

While challenges remain regarding planning permissions and grid capacity, the strategy of using data centres to export computing power rather than just electricity is gaining traction. This approach is particularly suited for model training tasks rather than latency-sensitive applications. The initiative represents a strategic effort to utilize natural resources in the global AI race.