Artificial intelligence is drinking the water of towns: why it affects you even if you’ve never heard of it
Have you ever wondered where the “brain” that answers when you ask something to an artificial intelligence program is? It is not floating in the air. It lives in enormous buildings full of computers called data centers. And those buildings have an enormous thirst for water.
This week, in the United States, there has been a public uproar that few people outside know about: thousands of people have taken to the streets to protest because these data centers are using too much water from their towns and cities. We’ll tell you about it in simple words.
What a data center is (explained easily)
Think of a data center as a gigantic library full of computers instead of books. Everything you do on the internet —watching a video, sending a message, asking something to artificial intelligence— is processed and stored in one of these buildings.
The problem is that these computers get extremely hot when they work. It is like when you run: your body’s temperature rises and you need to sweat so you don’t burn. Well, computers cannot sweat. They need the building to be cooled from outside.
And what are they cooled with? With water. A lot of water.
The figure that says it all
A large data center can use up to 5 million gallons of water a day (about 19 million litres). To give you an idea: that is what a city of between 10,000 and 50,000 inhabitants uses in a day. And a very large part of that water, around 80%, evaporates when it is used, that is, it is not recovered.
There are striking cases. In a county in California, a new data center could use 260 million gallons a year (almost a billion litres) from the Colorado River. In another city in Oregon, Google’s data centers are responsible for almost 40% of all the water that the whole city consumes.
Why artificial intelligence is more to blame than the rest
Okay, careful: this problem is not entirely new, but artificial intelligence has triggered it. Modern artificial intelligence programs need to work with lots of computers at once, all at the same time. The more computers work together, the more heat they make and the more water they need to cool down.
That is why the water and electricity consumption of these buildings is growing so fast. Companies are building new data centers at full speed, almost always where the land is cheap, even though there water is precisely scarce.
People get tired: protests in 42 states
In July there was the first day of coordinated protests in the whole country: about 142 demonstrations in 42 states at once. There were signs against big companies like Google, Amazon or Microsoft. The neighbours are not asking for artificial intelligence to disappear. What they ask is that they be asked before the water of their community is spent, and that the companies be held accountable.
In some places there are already people arrested because of the protests, at least 37 over the whole year. And some city councils, such as Seattle’s, have started to hold back the construction of these buildings until the water issue is clarified.
The most repeated example: a company used up in an area of Georgia 30 million gallons of water (about 113 million litres) before its first bill even arrived. That is, they used the water first, and asked later.
And what happens in Spain?
This debate is going to end up reaching Europe and Spain. Artificial intelligence is used the same way here, and the centers that support it need the same water. Spain has areas with little water and lots of heat, precisely where companies want to build these buildings. The experts already warn that the same conflict is brewing in several European countries.
You do not need to know anything about computing to understand the idea: every time we use a tool with artificial intelligence, behind it there are computers that need water. And that water comes out of someone’s tap.
The underlying question
The problem is not only how much each company uses. It is that water cannot be borrowed from one place to another. The carbon that a data center in one country pollutes does not matter where the wind blows. But the water that a data center in your town uses is water that your neighbours no longer drink. It cannot be brought from elsewhere.
Cooling technology improves and uses less water over time, but it costs money and many companies delay the change. Meanwhile, construction keeps advancing faster than the rules. The big question of the coming months is whether governments will set clear limits on the water these buildings can use, or whether every town will have to fight on its own.
Meanwhile, now you know: when you ask something to an artificial intelligence program, behind it there is water. Quite a lot of water.
