As droughts spread across Spain, the US, and elsewhere, communities are battling water-hungry data centers; Meta expects a €1B facility to use 665M liters/year
With drought spreading around the globe, battles over water are erupting between AI companies seeking more computing power …
Context & Ripple Effects
The water fight has been building for years: US communities first pushed back during the 2021 drought [[a:967611|when a typical data center was found to use as much water daily as a city of 30K-50K people]], and the Western megadrought then put hyperscale water use under sustained scrutiny in 2022. What changed is scale and geography — the EU now plans to at least triple AI data center capacity within five to seven years with water scarcity flagged specifically for Southern Europe, while governments elsewhere are already prioritizing compute over competing uses like housing as London's case shows.
Meta's €1B facility, projected at 665 million liters of water per year, lands squarely in that collision: droughts spreading across Spain and the US mean each new gigawatt-scale campus is now sited against a shrinking baseline, not an abundant one.
First-order effects
- Communities in drought-hit Spain and the US now face direct competition with Meta and other hyperscalers for municipal water supplies, turning local permitting hearings into the de facto gatekeepers of AI capacity.
- Meta's 665M-liter annual projection gives opponents a concrete number to organize around, raising the cost and timeline of every future site approval.
Second-order effects
- Hyperscalers will be pushed toward water-rich siting or expensive closed-loop cooling, shifting location economics away from cheap land and power alone.
- The EU's tripling plan runs into its own stated constraint: if Southern Europe's scarcity concerns harden into permitting rules, Brussels' capacity targets and its water reality pull in opposite directions, forcing operators to concentrate in Northern Europe instead.
Third-order effects
- Water joins power as a structural bottleneck for AI buildout — Big Tech's plans already face an estimated 44GW power crunch by 2028, and if water constraints bind similarly, aggregate AI capacity will fall short of announced ambitions regardless of capital available.
- Sustained community resistance points toward data centers being regulated like utility infrastructure, with mandatory water disclosure and allocation agreements becoming standard conditions of operation rather than voluntary ESG gestures.
The trend: AI infrastructure siting is shifting from a power-and-land problem to a power-and-water problem, with local water politics emerging as the binding constraint on where and how fast compute gets built.