Uncovering the Hidden Water Footprint of AI: Solutions for Quenching Its Insatiable Thirst

Our digital lives rely on data centres. But their cooling systems are threatening global water supplies, leaving citizens thirsty.

Author Kezia Rice, 06.16.25

When I travelled to Mexico in 2024, I spent weeks basking in the thick summer heat. But when I chatted with locals about the endless clear skies, they all told me they were praying for rain. Mexico was suffering from its worst drought since 2011. And 2024 was the warmest or second-warmest year on record throughout the whole of Latin America and the Caribbean.

Meanwhile, construction of data centres across Latin America has been booming; the market is expected to double by 2029. These data centres are being built by the likes of Google, Amazon and Microsoft to process the data transactions that power our digital lives. With the rise of AI, data centre demand is increasing even further. And because the supercomputers that run these transactions overheat, data centres require water to keep them cool.

Not just a bit of water. By 2027, AI’s annual water usage is predicted to be the same as the whole country of Denmark or half of the water used yearly by the UK. One data centre uses the equivalent daily water supply as a town of 10,000 to 50,000 people. And with the global demand for freshwater predicted to be 40 percent greater than supply by 2030, the planet doesn’t have water to spare.

‘Making AI Less “Thirsty”’

A 2023 study from the University of Colorado Riverside and the University of Texas Arlington shows that training ChatGPT-3 requires as much as 700,000 litres of freshwater—information that Microsoft kept under wraps. Read the full study here.

Data centres’ hidden water consumption

Much has been made of data centres’ energy consumption, so much so that Big Tech companies are all too keen to tell us when they’ve improved efficiency or reduced their carbon emissions. But this turned out to be not much more than greenwashing. Replacing electric fans with cooling systems might reduce carbon emissions but it also simply switches the burden from electricity to water.

According to Pablo Gámez Cersosimo, a researcher specialising in technology and biodiversity, “The water footprint of digitalisation could be larger and more problematic than its carbon footprint.” Add to this equation the fact that 49 percent of data centres don’t even record their water usage and we’ve got a massive problem we don’t even know the size of.

Green digital futures

How can we ensure a green digital future?

Growing e-waste, carbon emissions from AI, data centre water usage—is rampant digitalisation compatible with a healthy planet? Our latest project explores how digital tools and services can be developed with sustainability in mind.

The vicious circle of data centre construction

Big Tech companies present the construction of data centres in Latin America as an economic opportunity for the region. In reality, once constructed, data centres run almost autonomously and don’t boost economies as governments hope. Instead, they reduce drinking water supplies in countries which are increasingly affected by drought. The finished data centres release emissions, contributing to climate change and leading to more drought in a vicious, unending circle. 

But why are data centres depleting drinking water resources rather than using water that citizens wouldn’t drink anyway? It all comes down to cost. Contaminated water must be cleaned before use so it doesn’t corrode cooling systems––making fresh drinking water a cheaper alternative. While Microsoft has announced plans to begin reusing water in data centres from 2026, this will require more energy to cool down the water after it has been passed through the system.

With computer ownership concentrated in the Global North, another disparity plays out. In North America and Europe, our digital lives increase the demand for data centres. But the impact of building data centres is felt most by the Global South. By taking these resources, modern-day colonialism plays out in the digital sphere.

In Uruguay, campaigners fought for the right to fresh drinking water in 2004. Now, they’re using that legislation to take action against Google’s latest data centre construction which will consume an estimated two million gallons of water a day. Meanwhile, Uruguay is suffering from its worst drought in 70 years. In the words of Daniel Peña, a researcher at Montevideo’s University of the Republic, “[Google’s data centre] will give Uruguay virtually nothing and at the same time bring in its wake a set of serious ecological and social problems.”

Big Tech continues to build data centres in areas hit by drought

A 2025 investigation by non-profit Source Material and the Guardian revealed Google, Amazon and Microsoft will collectively increase data centres in areas of water scarcity by 63 percent. Read the full study here.

Renewable energy and tougher legislation could turn the tide

Is there a way to reduce data centres’ impact on people and the planet by mitigating their insatiable thirst? Potential solutions include cooling data centres with electricity powered by solar panels or hosting data centres in cooler countries that suffer less from drought, such as Sweden. New lines of code added to the Linux operating system have the potential to reduce emissions by five percent. And new research on photonic integrated circuits shows their potential in data centre applications to produce less waste heat, reducing strain on cooling systems. But with land for data centres going cheap in Latin America, it’s unlikely that Big Tech will decide to put people and the environment before profit anytime soon.

A positive step forward is that the EU has recently introduced legislation requiring data centres to report their electricity and water consumption. Experts are calling for similar legislation to be introduced worldwide. Meanwhile, researchers in the field emphasise the importance of “tracking and reporting AI’s water consumption”, as per their study ‘Making AI Less “Thirsty”’.

Data centre cooling systems are drinking up water at an alarming rate: as well as the two litres of water that you likely consume a day, your digital activity now consumes another three. As our digital demand takes bigger gulps from global water supplies, our planet is in danger of, quite literally, dying of thirst.

Public Interest AI: Artificial Intelligence for the Common Good Requires a Rethink

AI that serves the common good? To do so, it must be designed differently to its big tech siblings. Theresa Züger is researching this at the HIIG institute.

Photonic Integrated Circuits “Break New Ground”, Reducing Waste Heat of Data Centres

New study reveals how Photonic Integrated Circuits could reduce data centre energy consumption by replacing electrons with light particles.

Tomorrow’s AI, Today’s Problem: How Toxic GAI E-Waste Could Engulf the Planet

Generative AI (GAI) is taking over the world—in more ways than one. One new study has shed light on the hidden impact of LLM e-waste, in the hope that we can stem its tide before it's too late.

AI-Model Aardvark Forecasts Weather Without the Need for Carbon-Emitting Supercomputers

AI-model Aardvark can predict weather faster and more accurately than existing systems—all while emitting thousands of times less carbon.

The Hidden Gaps in Our Digital Carbon Footprint: A Conversation with Jens Gröger of Öko-Institut

Devices, data centres and our online lives all generate high CO2 emissions. Jens Gröger shares solutions for this digital carbon footprint.

Simplifying Scope 3 Decarbonisation with ClimateChoice 

Scope 3 emissions can make up to 90 percent of a business’s carbon footprint.  ClimateChoice helps companies understand and tackle them.

38C3 - Blick auf das CCC-Gebäude in Hamburg.
© Thomas Fricke
Hacker Congress 38C3: All About Sustainability in the Digital World

The Chaos Computer Club's annual conference took place at the end of 2024. Here are the top presentations to watch.

AI-Simulations Improve Aircraft Efficiency: The Solution the Aviation Industry Desperately Needs?

New software uses artificial intelligence to design more fuel-efficient aircraft. Will the technology sufficiently reduce aircraft emissions?