Water scarcity is increasingly shaping the global energy sector, forcing companies to rethink cooling strategies for power plants and data centres.
Research by Wood Mackenzie highlights the growing risk posed by high water stress, which could expose 31% of global GDP by 2050, up from 24% in 2010.
The report notes that thermal, nuclear, and hydro plants, which generated 80% of global electricity in 2025, depend heavily on water for cooling. In Europe, recent heatwaves and low river flows prompted temporary reactor curtailments, underscoring the sector’s vulnerability.
Experts warn that AI computing is accelerating demand for liquid-cooled data centres, which require water volumes far beyond air-cooling capacities.
“AI clusters generate heat loads that air simply cannot handle at scale,” said Jom Madan, principal analyst at Wood Mackenzie. “Liquid cooling isn’t optional anymore. It’s the foundation for next-generation compute. The water question hasn’t gone away; it’s moved from the data hall to the power plant.”
Traditional wet cooling towers reduce water withdrawals but triple consumption through evaporation, while dry cooling eliminates water use entirely at the cost of a 7-percentage-point efficiency penalty and higher capital expenses. Countries including India, Mexico, Egypt, and Turkey account for over half of the GDP exposed to water stress.
The research also highlights the impact of AI on data centre design. Modern training clusters often exceed 120–200 kilowatts per rack, far surpassing air-cooling limits. Hyperscale operators are adopting liquid systems capable of handling up to 250 kilowatts per rack, which improve energy efficiency but raise water use, pushing water usage effectiveness up by an expected 20% by 2028.
Recent reporting shows that these water challenges are already materializing. In Australia, municipal drinking water systems are under pressure from major facilities whose cooling demand rivals urban consumption, The Guardian reported.
Analysts in the US have warned that the AI boom is colliding with “an unplanned water problem,” as data centres expand in water-sensitive areas, a Forbes report noted. A mid-sized data center may require about 300,000 gallons of water per day for cooling.
By 2020, cooling systems accounted for roughly 40% of total data centre electricity usage. Rising energy costs and grid constraints are also cited as potential limits on big-tech AI investment plans, with planned spending of approximately $635 billion facing headwinds linked to cooling and water availability, according to Reuters.
Madan added that while renewable energy sources like wind and solar reduce water dependency, policy frameworks are needed to accelerate adoption of hybrid and dry cooling technologies at the pace the market demands.
Wood Mackenzie tracks 250 GW of data centre projects globally, with power usage effectiveness expected to reach 1.2 by 2028.
