A ‘thirsty’ AI boom could deepen Big Tech’s water crisis

Let’s say I ask 10 to 50 questions of 

ChatGPT. The hugely popular chatbot would need to gulp about 500 milliliters 

of water to provide me with the answers. That’s roughly the amount in 

this standard bottle of water. Hundreds of millions of users then, all submitting 

prompts quickly shows then just how thirsty artificial intelligence can be. It’s little-known details like this that help to shine a light on Big Tech’s 

water footprint, an often-overlooked issue, but one that is increasingly coming under public 

scrutiny as global demand for AI gathers pace. In July of 2023, for example, protesters took 

to the streets of Uruguay’s capital city to demonstrate against Google’s plan to build a data center. The proposal sought to use vast quantities of water at a time when the South American country 

was suffering its worst drought in 74 years. Shaolei Ren, a researcher at the University 

of California, recently published a paper looking at the resources needed to power AI 

models — as well as what can be done to try to quench the thirst of this juggernaut. In general, the public is getting more knowledgeable and aware of the water issue.

If they learn that the Big Tech’s are taking away their water resources and they are 

not getting enough water, nobody will like it. I think we are going to see more clashes 

over the water usage in the coming years as well so this type of risk will have 

to be taken care of by the companies. In just the last few months, tech giants 

deeply involved with AI research have reported a massive jump in their water consumption. 

Researchers say that AI is partly to blame, as companies use millions of liters of water to 

keep temperatures in check at their data centers. It’s easy to think of the internet as something 

that’s intangible, but whether it’s asking questions of ChatGPT or logging in to check 

your emails, the computing power needed to keep these websites running relies on a lot of water. Microsoft disclosed in its latest environmental sustainability report that its global water use 

rose by more than a third from 2021 to 2022, climbing to nearly 1.7 billion gallons.

For Google, total water consumption at its data centers and offices came in at 5.6 billion gallons 

in 2022, a 21% increase on the year before. It means that Microsoft’s annual water use was 

the equivalent to more than 2,500 Olympic-sized swimming pools, while for Google, this would 

be enough water to fill nearly 8,500 pools. Both companies are working to reduce their 

water footprint and become “water positive” by the end of the decade, meaning they 

aim to replenish more water than they use. It’s notable, however, that their latest 

water consumption figures were disclosed before the launch of their own

respective ChatGPT competitors. The computing power needed to run Microsoft’s Bing Chat and Google Bard could 

mean more water usage over the coming months. Water scarcity is a big problem 

on every continent — and one that is being exacerbated by the climate crisis.

Here in Dubai, a region where heat is extreme and water is precious, policymakers and 

business leaders are gathering at COP28 to discuss what more can be done. What really needs to be in place is an international governance body that can regulate, 

and that is what COP is to a certain extent, a global governance institution that tries to 

regulate and tries to monitor complex problems. One subset of that complex problem is AI, 

which is at the moment in a very transient phase. Its an inflection moment for AI so it 

could go either way and it is very important that we govern that direction to be less energy 

intensive and more within planetary boundaries. Data centers are part of the lifeblood of Big 

Tech — and a lot of water is required to keep them running smoothly. For Meta, its these data centers 

that generate not only the highest percentage of its water use but also the lion-share of 

its energy use and greenhouse gas emissions. Rows upon rows of computing servers in hubs the 

size of warehouses work to store and transfer all our digital information, with almost 

all this energy then converted into heat. Companies like Google, Amazon, Meta and 

Microsoft all use various cooling systems to prevent their servers from overheating. 

One common method involves the use of cooling towers. This heat-rejection device 

works like a giant air conditioner, typically using water in an evaporation process to 

expel building heat into the atmosphere. Another technique often used is outside air cooling — or 

“free” cooling. This system takes the outside air, which has a lower temperature than inside the 

building, to directly cool down the servers. This process can work without the need for 

cooling towers, but water evaporation is still needed when the outside air is too hot, 

often at the height of summer when people use water the most and drought is at its worst. Basing data centers in cooler climates is one way to get this free cooling but, at present, 

only a fraction of them are currently housed in colder environments. The U.S., by a long 

way, operates far more data centers than any other country, with more than 5,300. That’s 

10 times more than second-placed Germany. Ann Skeet, senior director of leadership ethics 

at the Markkula Center for Applied Ethics, says tech companies will need to 

be a little more judicious on AI if they are to meet their environmental goals. I think we are going to see some activity from regulators. There is already some increased 

requirements coming in the new year for companies to disclose their water usage in the 

European Union. And I think that requirement is a necessary one for people to start to appreciate 

the issue and think about their own choices in it. So, I think they have to become a little 

more heavy-handed in terms of what they are asking companies to do and even think about 

placing certain limitations on companies if we don’t see either market forces or innovation 

contribute to bringing the use of water down. AI is one of most prominent and 

fastest growing workloads in data centers, which means Big Tech is likely to need 

a lot more of this precious resource. In fact, the computing power required for AI 

is estimated to double every 100 days and it is expected to increase by more than a 

million times over the next five years. Ren and his colleagues have predicted 

that global AI demand could be accountable for up to 6.6 billion cubic meters of 

water withdrawal by 2027. Remarkably, that would be roughly equivalent to the total 

annual water withdrawal of Washington state. Researchers say this is deeply concerning at a 

time of growing water scarcity. Alongside the urgent environmental risks, tension over the water 

usage of data centers and human needs has been at the heart of social problems across the globe. Following the protest in Uruguay over Google’s plan to build a data center, the tech giant 

reportedly said the project was still at an exploratory phase and stressed that sustainability 

remained at the heart of its mission. It’s thought social challenges such as this 

are likely to increase over the coming years, prompting many to ask what some of 

the solutions could be to meaningfully address Big Tech’s growing water footprint. This water is a shared challenge by all of us. So, nobody can solve this issue without global 

efforts. So, as users, if, let’s say we are just doing some random stuff, we might want to avoid using 

generative AI because those are generally more resource-consuming compared to a simple search. So there are a lot of ways that individuals users can

contribute to the water usage saving, But the first step is that we need to be aware of this issue or we need to have the visibility 

of the water efficiency of the services.