The rapid expansion of artificial intelligence is driving an unprecedented surge in energy and water consumption by data centres worldwide, with a single large facility now using up to 300,000 gallons of water per day for cooling, according to a new analysis. This hidden environmental cost threatens to undermine the benefits of AI unless urgent action is taken.
Data Centre Growth Outpaces Efficiency Gains
Global data centre electricity consumption reached 460 terawatt-hours in 2025, equivalent to 2% of worldwide electricity use, and is projected to double by 2030. Water usage follows a similar trajectory, with the average hyperscale data centre consuming 1.2 million litres daily. In drought-prone regions like California and Arizona, this strain on local water supplies has sparked protests and regulatory scrutiny.
“We are trading one environmental problem for another,” said Dr. Emma Green, an energy policy researcher at the University of Sydney. “The computational benefits of AI come at a steep ecological price that is often overlooked.”
Cooling Systems Drive Water Demand
Most data centres rely on evaporative cooling systems that consume large volumes of water. In 2025, data centres in the United States used an estimated 200 billion litres of water, enough to supply 200,000 households for a year. Newer direct-to-chip liquid cooling technologies can reduce water use but increase energy demands, creating a trade-off.
Google and Microsoft have pledged to become water-positive by 2030, but current trends show water consumption rising by 20% annually. Facebook’s parent company Meta reported a 34% increase in water use in 2025, driven largely by AI workloads.
Renewable Energy Not a Panacea
While many tech giants purchase renewable energy credits, data centres still draw power from grids that rely on fossil fuels. In Virginia, the world’s largest data centre hub, electricity demand has grown so fast that utility companies have delayed coal plant retirements. “Renewable energy alone cannot solve the water problem,” noted Green. “We need systemic changes in how we design and operate these facilities.”
Policy and Innovation Needed
Governments are beginning to respond. The European Union’s proposed Data Centre Energy Efficiency Directive would require facilities to report energy and water use and meet minimum efficiency standards. In Australia, the New South Wales government has imposed a moratorium on new data centre developments in water-stressed areas.
Innovations such as liquid immersion cooling, which submerges servers in non-conductive fluid, can cut water use by 90% but remain expensive. Meanwhile, AI itself can help optimise cooling systems, reducing energy use by 15-30%.
Balancing Progress and Sustainability
The challenge is to harness AI’s benefits—from medical diagnostics to climate modelling—without exacerbating environmental crises. “We need a holistic approach that includes efficiency standards, water recycling, and location planning,” said Green. “Otherwise, the digital revolution will leave a very physical scar.”
