Will AI Data Centres Drive Up Australian Electricity Bills?

The analysis projects that AI data centres could consume 34 TWh of electricity annually in Australia by 2035, equivalent to a quarter of current national demand. The article examines whether this load will push up household power bills, what regulatory frameworks apply, and how consumers might be protected. No specific wholesale cost impact or bill forecast is offered in the article.

Where this fits

This demand spike is a direct stress test on Australia's renewables-only transition plan. A 34 TWh load arriving in 13 years must be met by new generation, and the grid will depend on what that generation is. If met by weather-dependent solar and wind, it will require proportional firming capacity (gas plants or batteries) and a transmission overbuild to handle intermittency and congestion. That entire stack, renewables plus firming plus storage plus transmission, costs more per unit of delivered power than firm generation. A nuclear plant of equivalent firm output would occupy a fraction of the land, require a fraction of the critical minerals (lithium, cobalt, nickel, processed rare earths from China), and produce power independent of weather. Instead, Australia is choosing the renewable-plus-gas path, which locks in higher whole-system costs and sustained reliance on gas. The data-centre load makes this choice visible in real time.

What it means

A 34 TWh AI load is a decarbonisation test case. Australia can meet it with renewable intermittency and gas firming, or with firm low-carbon generation. The cost difference is material. On grid resilience, weather-dependent supply of a strategic load (data centres serve international tech firms) introduces a new vulnerability: a renewable grid under drought or low-wind conditions must either shed AI load or import power at peak prices, or both. Sovereignty is the third pressure: the minerals, inverters, and battery cells to firm 34 TWh of renewable output will be sourced from China.