Australia Lacks Data on AI Data-centre Water and Power Use
In The Conversation, Michael Vardon argues that Australia lacks the granular, public data needed to judge how much water and electricity new AI data centres will use. The article notes OpenAI's Sam Altman has said Australia could become a "data centre capital of the world." Citing reported figures, Vardon estimates Australia's roughly 300 existing data centres (about 1.3 gigawatts) consume on the order of 15,000 to 35,000 megalitres of water a year, a small fraction of one percent of national water use. He highlights that a megalitre of water is worth far more in a data centre (about $2.3 million) than on a farm, roughly 500 times more. The piece concludes that energy, not water, is the larger systemic risk and calls for site-level measurement before major planning decisions are finalised.
What happened
In a piece for The Conversation, Michael Vardon argues that Australia currently lacks granular, public data on the water and electricity consumption of newly proposed and built AI data centres. The article notes that OpenAI chief executive Sam Altman has said Australia could become a "data centre capital of the world." Using reported figures, Vardon estimates the country's roughly 300 existing data centres, with about 1.3 gigawatts of capacity, consume on the order of 15,000 to 35,000 megalitres of water per year, a small fraction of one percent of national water use.
The economics
The article emphasizes that water is far more economically valuable in a data centre than in agriculture: about $2.3 million per megalitre, roughly 500 times the value of the same water used on a farm. Vardon argues this density does not by itself resolve trade-offs with agricultural or ecosystem water needs, and that public per-megalitre comparisons can mislead without site-level detail.
Why it matters
Class B analysis: measuring the environmental footprint of compute-heavy facilities requires site-level telemetry, cooling-system water accounting, and energy-source disclosures. Operators in comparable markets track metrics such as power usage effectiveness and metered coolant flows, but those figures are rarely public at fine granularity, which is the gap the article highlights.
What to watch
- •Emergence of site-level disclosure requirements for water and energy reporting in Australia.
- •Grid-supply studies quantifying incremental peak and annual demand from AI clusters, and whether new gas capacity is proposed to meet it.
- •Whether energy, rather than water, remains the dominant constraint as more capacity is built.
Scoring Rationale
This is a timely infrastructure and policy analysis on a high-salience topic, the resource footprint of AI data centres, and it usefully reframes water as currently minor while flagging energy as the systemic risk. It is an opinion-and-analysis data-availability argument rather than a technical breakthrough or hard-news disclosure, which caps its impact.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
