Are data centers fueling the global water crisis?

DW (English)

In Chile, Uruguay, and the United States, local communities are protesting against tech giants like Google and Amazon for their artificial intelligence technology and its growing thirst for water.

Transcript

00:00Protests against tech giants are erupting everywhere, in Chile, Uruguay and the United

00:07States. While water pipes run dry and the soil dries out, Amazon, Microsoft and Google are

00:15building massive data centers in regions already suffering from drought. Computing power requires

00:21enormous amounts of water, leaving little for local communities and agriculture. More data,

00:27more cloud computing, more AI. And it's accelerating with no signs of slowing down. Environmental

00:34protection or technological progress? Is it possible to have both? Data centers and the battle for water.

00:41Typically a data center is planned and then there's a lot of pushback. There have been protests

00:50in Latin America. The region is facing extreme drought and water shortages. Uruguay especially

00:56struggled in 2023 during its worst drought in 75 years. And then Google happened to show up at

01:06exactly the same time. Google wanted to use 7.6 million liters of water per day,

01:14enough to supply 22,000 people. And during a drought, an extreme water crisis.

01:19Daniel Pena is a scientist and activist. He's critical of the lack of community input.

01:27There was no public dialogue and no public hearing. The project was classified as category B,

01:35considered low to moderate risk. As a result, public involvement was minimal.

01:41Daniel Pena is concerned about significant environmental impacts from the Google Cloud

01:48project. The Santa Lucia River is important for the capital city Montevideo. The battle over water

01:55has already begun here. When there's no clean water from the tap, bottled water is the only option left.

02:01Not everyone can afford bottled water. People on low incomes don't have the money.

02:11The water tastes bad. It causes stomach aches. Sometimes even headaches.

02:19AI and cloud computing are booming. Data centers, the backbone of AI,

02:24need massive computing power and water. Global demand is growing exponentially.

02:32These huge centers are often built in some of the world's driest regions. And resistance is growing.

02:38People want more transparency, a say in the process, clean air and clean water.

02:44That's the situation in Uruguay in the summer of 2025. Construction on Google's 850 million dollar

02:51cloud project near Montevideo has been underway for a year and will take more than two.

02:56It promises jobs and innovation and local and national leaders see it as a key investment.

03:06I think the process was good. I think it's clear that it's now in the construction phase

03:13and that the project hasn't caused any major inconveniences.

03:16We hope that once it's up and running they can generate synergies that can attract other types

03:28of industries and activities associated with technological innovation.

03:35What companies are we talking about? Who are the major global players in the computing power business?

03:40It's the big three. Amazon subsidiary, AWS, Microsoft Azure and Google Cloud. Together they account for two-thirds of the market.

03:51Why is so much being built? Because we want it. Industry, the economy, society. The demand is there.

03:58But the environmental impact is significant. In many large data centers much of the cooling water simply evaporates.

04:05That's because it's not reused in a closed loop but sent through cooling towers where it absorbs heat from servers and escapes as vapor.

04:20In fact evaporation is the most efficient way of removing heat. That's why we for example we sweat.

04:27When we get too hot that evaporation of the of our sweat off of our skin is an effective way of keeping us cool and keeping us from overheating.

04:35Aron Wemhoff is a data center sustainability expert. Google alone is now said to consume 20 billion liters of water annually in the US.

04:46That's a lot of bathtubs. But let's put that in Olympic sized swimming pools. 50 meters long that would be 8000 pools.

04:56I do know several colleagues at Google and I know that Google is actively trying to do things to mitigate their water footprint.

05:04Worldwide data center water consumption is increasing drastically and is expected to more than double by 2030.

05:11In fact following recent protests Google has spoken of a commitment to responsible water use

05:17and now plans to cool its data center with air instead of water.

05:23Sounds like a simple solution but it's not.

05:26Water cools more effectively due to its physical properties.

05:30Air cooling ultimately requires more energy.

05:34And environmentally friendly electricity is often not sufficiently available for data centers.

05:40That's when coal and gas power plants come into play.

05:42So maybe water cooling after all?

05:45It is possible or likely that many waterless cooling systems are less energy efficient

05:53compared to systems that utilize evaporation.

05:56And as a result you end up just having a higher water footprint upstream at the power generation source

06:03as opposed to at the data center site itself.

06:06We've seen this before. In Chile for instance where Google's proposed data center drew public backlash

06:13over concerns about excessive water use. Santiago's environmental court partially revoked

06:18the project's approval citing insufficient consideration of groundwater and climate impacts.

06:23Now Google has to revise its plans and will likely switch to air cooling.

06:28So why build so many data centers in hot, dry regions with scarce water?

06:34The main reason for looking for drier areas is both an efficiency reason and an avoidance of corrosion.

06:44The humidity can't be too high, warns researcher Lorena Jaume-Palassi,

06:48or the costly equipment will begin to corrode.

06:51If you use something else than drinkable water to cool,

06:55if you are using liquid cooling. So this means that you need drinkable water to avoid corrosion

07:02to cool down your data center and you enrich that water with some sort of chemical stuff.

07:10So this means you cannot bring back that water into the normal water circuit.

07:15So conflicts in Chile as well. But what could be done better when choosing locations?

07:20I will say that my opinion as being somebody who is pushing for minimizing the environmental impact

07:27of data centers, I would like to see more data centers being built near large bodies of water

07:31in cooler climates. Let's turn to Northern Europe, specifically Sweden. In Falun, about 200 kilometers

07:38from Stockholm, there's a remarkable data center pushing the boundaries of design and sustainability.

07:43It's all about resource use, water, electricity and more. In Falun, there's no shortage of water.

07:51It's drawn from a nearby lake, treated, stored in tanks and sprayed to cool the system using the

07:56principle of evaporation. The server rooms themselves are also cooled with water,

08:02much of which is reused where possible. Large air ducts blow cold air into the server rooms.

08:09This uses electricity. But here, it's 100% renewable. Sweden relies on hydroelectric power,

08:16both large and small scale, and wind energy also supports the data center.

08:21And the processor heat is not just released into the environment.

08:28Some of the water, of the reused heat, is distributed to a district heating plant just

08:35as a neighbor of us, where we supply them with heat in their manufacturing of the wooden pellets for furnaces.

08:49Pierre Wellander is the Falun site manager. The local energy provider is just across the street.

08:55Pellets are dried and burned in a small plant that supplies district heating.

08:59Sweden is well suited for data centers. There's plenty of water, wind, and its northern location helps too.

09:09During winter time, we have cold temperature. We have like 20, normally 20 degrees below zero,

09:16which gives us free cooling, actually.

09:19So just let the cold air in. It might not be entirely fair to compare Sweden with Chile or Uruguay,

09:26but some ideas might still be transferable. Take wood for example. In Sweden it's a sustainable,

09:32eco-friendly building material that binds carbon dioxide.

09:38Now we're up to 95% of wood. And we find out that this is the most perfect way of building a computer center.

09:46Back to the point of community input. Falun faced similar concerns as in Uruguay and Chile.

09:52And through that whole process, there has been a great focus on having a dialogue with the local

10:01population. The people of Falun has been offered to pose and ask any kind of question. And of course,

10:13they've also been answered when asked those questions in a transparent and open way.

10:21Overall, northern countries offer more pros than cons for data centers,

10:26even with factors like higher humidity. At 5,427 the US has the most data centers,

10:34followed by Germany at a distant second. In Latin America, there are relatively few.

10:40Another one just opened in Memphis, Tennessee, a data center for artificial intelligence.

10:46City officials see it as the future, but the locals are furious. Especially after learning Elon Musk and

10:53his chatbot Grok are behind it. Progress, yes, but not at people's expense. It was built in just 122 days,

11:03and it's still growing. The process and how this business transaction took place was without the

11:14knowledge of myself and other council members. I found out on the news. Water is life. This is an

11:25absolutely critical resource that everything needs for survival. Industries do need water for their

11:32process. Businesses do need water to make their revenue, but that should never be put in front of

11:38the drinking water needs of residents. People are raising concerns about air and groundwater pollution

11:44and ignored environmental laws. Gas turbines are running to generate electricity, producing emissions,

11:51which, they say, didn't initially have the proper permits. Once again, too little information for

11:57those affected. Perhaps a textbook case of how not to do things. Tech giants face accusations of lacking

12:04transparency. Amazon, Microsoft and Google are withholding key data on water and electricity use.

12:10Google also declined our request to visit their site in Uruguay. At this point, we cannot facilitate a

12:17visit to the site nor an interview. Reliable data on electricity and water is scarce. Most of these other

12:25aspects of water consumption are not being indexed, are not being measured, so it's really very

12:32intransparent. I have my doubts, yes, because until now what we can only do as researchers is try to

12:40use open source. We really don't know how much water they really consume. Water consumption is already high

12:49and projected to double, but data center electricity consumption will also increase by 128 percent by 2030.

12:57We don't know how chip performance will evolve. Will efficiency gains keep up with ever increasing demands?

13:04The Chinese AI chatbot DeepSeq made headlines for needing less computing power to generate responses.

13:12It reportedly uses far less energy than comparable models from OpenAI, Microsoft or Google. Perhaps

13:20future electricity and water demands won't be as high as predicted. Amazon, Microsoft and Google say their

13:27goal is to be water positive by 2030, meaning they'll provide more water than they use. The tech giants have

13:35a clear interest in being efficient and saving energy, but how water positivity will work and whether it's

13:41achievable remains unclear and controversial. High water and energy use, that's likely the price of progress,

13:50but the environment bears the cost. Small steps can help reduce the computing load,

13:55making less resource-intensive videos, using a calculator instead of AI for simple math,

14:01or just using a search engine instead of asking AI.

14:07Data centers are coming, but people want to know what that means for them. Environmental protection

14:12and data production still don't go hand in hand. What's more important to you? Unlimited progress or

14:19the sustainable use of resources? Let us know in the comments.

Category

Transcript

Be the first to comment

Recommended