Water Is Becoming the Next Gatekeeper for Europe's AI Data Centers
A Bluefield Research projection puts water related spending at €6.8 billion by 2036 as cooling demand, permitting rules, and water stress mapping collide with the AI buildout.
A Bluefield Research projection puts water related spending at €6.8 billion by 2036 as cooling demand, permitting rules, and water stress mapping collide with the AI buildout.
By 2036, Europe's AI data centers will have spent roughly €6.8 billion on water systems, according to a new industry forecast, putting water alongside power and chips as a siting constraint rather than a utility line item.
The projection comes from market analysts at Bluefield Research and covers the period 2026 to 2036, with the company's analyst Zineb Moumen framing water as a design, operational, and permitting constraint rather than just a recurring cost. About 62 percent of that spend is capital expenditure: cooling systems, water treatment and reuse infrastructure, municipal water connections, and pretreatment facilities; the upfront build-out decisions that determine whether a site can sustain AI workloads at all. The remaining 38 percent is operating expense: ongoing water consumption, treatment chemicals, and service contracts.
The €6.8 billion is a baseline scenario. Bluefield's range for the same report runs from roughly €4.7 billion in a slower AI buildout to about €7.1 billion on the upside, with the variance driven by AI capex pace, liquid-cooling adoption, and how quickly European permitting clears new sites. Treat the number as a projection under stated assumptions, not a guaranteed spend plan.
Three countries carry roughly 40.5 percent of the projected regional spend: Germany, the United Kingdom, and France. Hyperscale footprints, the largest and most power-hungry class of data centers, have already clustered there, and those utility and permitting regimes can still, for now, host large campuses. That concentration is also where water stress is loudest in Europe, which is why the next wave of AI sites will increasingly be steered by sub-national water mapping rather than national AI strategy alone.
Behind the direct-water number sits a second-order load. Power plants cooling the electricity feeding data centers are projected to drive about a 30 percent rise in indirect water consumption over the forecast window, on top of the facility-level figure. Liquid cooling, forecast at roughly 20 percent of new European deployments by 2036, trims direct water intensity but shifts the burden. Heat rejection moves to larger cooling towers or district loops, each with its own permitting and water-rights exposure.
New EU corporate water disclosure rules are forcing hyperscale operators to publish facility-level water use for the first time, and trade-press coverage of the Bluefield forecast points to the same convergence: disclosure, water-stress mapping, and tightening local permits are now running in parallel with AI demand rather than after it. That is the moment analyst projection and site-selection reality meet.
The next trigger worth watching is how Germany's permitting authorities handle the first hyperscale water reuse filings under the new disclosure regime, a decision that will test whether the €6.8 billion projection doubles as a geographic shift in Europe's AI map, or stays a forecast.