AI’s hidden thirst: the water and power cost of data centres

Data centres worldwide consume vast amounts of electricity and water, potentially straining resources. With AI’s growth, the footprint is set to balloon.

Vaishnavi Chandrashekhar

Pic credit: Wikimedia Commons

Last month, a UK government agency called on residents to help mitigate the country’s water crisis.  Among the Environment Agency’s suggestions on how people could save water were expected common-sense ones such as turning off taps and fixing leaking pipes. And then there was this advice: delete old emails and pictures. 

Wait, what?

If that was your reaction, join the company. The agency raised some eyebrows - and a few laughs — with its suggestion. And at first hearing, the advice does sound absurd: surely emails have nothing to do with water use.  

But it’s not that simple, it turns out.  

Obviously, emails—or any of our other expanding digital footprints— don’t directly consume water. But emails, WhatsApp and the other bits and bobs that make up our digital economy are stored in data centres that guzzle electricity to run the servers and routers and then water to cool them down. 

How much water and energy?

In the US, data centres are estimated to collectively consume 1.7 billion litres of water a da. Google reported that its data centres worldwide consumed around 30 billion litres of water in 2024 alone. A large data centre can consume a small town’s worth of water.  

As for electricity, data centres consumed around 1.5% of the world’s electricity in 2024, according to the International Energy Agency (IEA). By 2030, the agency projects, this share will grow to 8%. As our digital world grows, so too will data centres. 

But most environmentalists aren’t worried about regular emails—much of the projected increase in data centre energy consumption will come from the explosion of AI systems. That’s because AI systems are much more resource intensive. One commonly cited estimate suggests a ChatGPT query consumed 10 times more energy than a regular Google search. 

AI-focused data centres can draw as much electricity as power-intensive factories such as aluminium smelters or 100,000 households, according to the IEA. “The largest ones under construction today will consume 20 times as much],” it says. 

And because these data centres tend to be geographically concentrated, they can put pressure on local power and water supply. Nearly half the data centre capacity in the United States, the largest consumer of data centre energy, is in five regional clusters. The state of Virginia, for instance, has 340 data centres that are estimated to account for more than a quarter of the state’s electricity use. Local communities are worried.

In India, there are some 262 data centres, the seventh most in the world, with more expected to come up as investment in the sector grows. Most of the centres are clustered around the big metros—Mumbai, Hyderabad, Delhi, Bengaluru, and Chennai—which makes for efficiency but could burden local grids and water supplies. Rural communities around the country already face inadequate power and falling groundwater levels, and the growth of this new industry could add to that problem. 

A report published in The Guardian earlier this year found that the big tech companies—Amazon, Google, Microsoft—operate data centres in some of the world’s driest areas. “Amazon’s three proposed new datacentres in the Aragon region of northern Spain – each next to an existing Amazon datacentre – are licensed to use an estimated 755,720 cubic metres of water a year, roughly enough to irrigate 233 hectares (576 acres) of corn, one of the region’s main crops,” the report said. 

Still, there’s reason to be optimistic. Some of these companies are signalling moves to reduce their water and electricity consumption. Microsoft is planning a “zero water” datacentre. 

As for energy, the IEA expects half the growth in data centre demand to be met by renewable energy, which is also expanding. Many point out that AI can also help save energy and reduce carbon emissions by making industrial and logistical systems and processes more efficient. This reduction could end up cancelling out the growth in data centre emissions.  (Or it might not: One fear is that gains in efficiency will be offset by overall growth in consumption.)

There are also signs that chatbots are becoming more efficient. While estimates from last year said an AI chatbot search uses ten times more energy than a regular Google search, more recent information suggests the difference may have shrunk. (AI firms aren’t always transparent.) If the energy gains are true, a Washington Post report recently said, “[A]asking AI eight simple text questions a day, every day of the year, adds up to less than 0.1 ounces of climate pollution. That’s 0.003 percent of the average American’s annual carbon footprint.”

In India, researchers at the Institute for Energy Economics and Financial Analysis point out, data centres are “prime sites for renewable energy resources such as rooftop solar, battery + energy storage systems (BESS) and local microgrid.” The biggest of the India centres, near Mumbai, is said to get half its power from renewables. 

As for the water needed for cooling, they suggest that locating new centres near coastal areas would allow them to use seawater for the purpose—and this water does not have to desalinated for use. (Some companies seem conscious of the issues: Google is reported to be investing up to $6 billion to establish a 1-gigawatt data centre powered by renewables in Andhra Pradesh.) However, experts have noted that companies tend to prefer inland locations for data centres due to cheaper land prices and possibly less corrosion from salty air. Coastal locations could also lead to other issues if environmentally sensitive land is used or the industry’s water consumption affects traditional livelihoods. 

It sounds like some rules are needed for the sector. In June, UNEP issued guidelines to help governments reduce the energy and water consumption of data centres, in a bid to fill a gap in regulations and best practices. 

Few countries have rules to ensure data centre operations are truly sustainable, the agency noted. “We know that data centres consume large amounts of energy and water, and we know that consumption is only going to grow, which will mean more greenhouse gas emissions and greater stress on water supplies,” said Martin Krause, Director of UNEP’s Climate Change Division. 

UNEP’s guidelines, which were reviewed by service providers as well as researchers, include recommendations for countries on power usage effectiveness, water usage effectiveness and renewable energy use in selecting equipment and facilities for data centres. They suggest the incentive mechanisms like labelling schemes to encourage data centre operators to meet sustainability standards. “The guidelines are crucial for developing countries, where expansion is most prevalent,” said UNEP. “As these nations expand their digital infrastructure, adopting sustainable practices is crucial to prevent the escalating consumption of energy and resources.”

So, does all this mean that UK’s Environment Agency advice to delete your mails to save water was correct? Perhaps not precisely at that moment. One technology reporter noted that England’s data centres are mostly located in the London area, not in the drought impacted northern regions of the country. 

But it’s still a good idea to think about our digital habits. In the real world, it’s easier to understand the environmental consequences of our actions---the energy and resources consumed by taking a bath, putting on the air-conditioner, or driving to work. But the impact of our actions in the digital world is almost invisible, certainly intangible in that moment of action. Yet, they do have an impact—in a building thrumming with servers far away.  

So, while you may not have to delete that old email or picture, you might still want to think for a moment before generating that Studio Ghibli-style image of Kuch Kuch Hota Hai or asking ChatGPT that random question. 

Earth Shifts, a monthly column on The Migration Story, will analyse the impact of global green goals amid mounting climate uncertainties on lives and livelihoods. 

Vaishnavi Chandrashekhar is an environment and science journalist based in Mumbai.