Developments · · 2 min read · Source: Greenpeace / Öko-Institut
Germany's data centers could double their water use by 2030
German data centers used about 13 billion liters of water for cooling in 2025, according to a new short study by the Öko-Institut. With the AI build-out, demand could reach up to 27 billion liters a year by 2030, says the analysis commissioned by Greenpeace.
Related tool · AdJulius – AI data analysisTry it free
Key facts
- German data centers used about 13 billion liters of cooling water in 2025, according to the Öko-Institut.
- Demand could reach up to 27 billion liters a year by 2030; around 30 projects of 100 MW or more are planned.
- Only 139 of 319 published operator reports were plausible, the study says; Bitkom disputes parts of the calculation.
The key numbers
The Öko-Institut analyzed data from Germany's national data center register. Here's what it found:
- 13 billion liters of cooling water in 2025, roughly the volume of Berlin's Wannsee lake.
- Up to 27 billion liters a year by 2030, as the study expects IT power use in data centers to grow from 19 to at least 55 terawatt hours.
- 0.67 liters per kilowatt hour of IT power is the weighted average, ranging from 0.35 to 1.07 liters depending on facility size.
- Around 30 large projects of 100 megawatts to more than a gigawatt are planned in Germany. A single 100 MW site can use 200 to 930 million liters a year, depending on its cooling.
Patchy data
Of an estimated 1,200 data centers required to report, only 484 operators submitted figures, and just 139 data sets were plausible enough to use. So the totals are an extrapolation from a narrow base. Germany's digital industry association Bitkom pushes back, saying about two thirds of facilities use no more than 0.1 liters per kilowatt of connected capacity and that significant consumption mainly comes from evaporative cooling.
Efficiency isn't the same as saving water
One finding stands out: energy-efficient data centers are not automatically water-efficient; the authors found no statistical link. With dry or compression cooling, direct water use can even drop to zero. Greenpeace is calling for water impact checks before permits, no water-intensive sites in dry regions and penalties for operators who ignore reporting rules.
What this means for you
- AI has a physical footprint: alongside power, increasingly water, and it's local.
- The data is thin: fewer than half of operators report at all.
- For companies, where and how their cloud is cooled is becoming a sustainability-reporting question.
How to use this
When choosing cloud or AI providers, ask for data center location, cooling technology and published water and power metrics. For routine tasks, a smaller model is often enough.

Luigi's take
For most professionals, AI is a tab in the browser. This study is a reminder that every prompt runs in a building that needs cooling. That doesn't mean you should stop using AI. But the build-out is happening now, in Europe too, and the debate over where and how to build will get louder.
My advice to companies: when you choose cloud or AI providers, ask about location, cooling technology and published metrics. If your sustainability report mentions AI, you should be able to answer those questions. And on a practical level, not every task needs the biggest model. A smaller model for routine work saves money and resources.
Source
Greenpeace / Öko-Institut: Kurzstudie: Wasserverbrauch deutscher Rechenzentren (Stand 11/2026, PDF)
More sources
- heise online – Bericht (dpa) mit Einordnung durch den Branchenverband Bitkom
- t3n – Bericht zur Studie
This article was researched and written with AI assistance. Responsible: Luigi Lugmayr.



