EXPLAINER: Why is AI so thirsty?

Artificial intelligence can do all sorts of things. Research. Write. Plan. Prioritize. It can be incredibly helpful if you have a thirst for knowledge. 

But AI has its own thirst. It runs on high-powered servers in data centers. The servers can heat up quickly, so these centers need cooling systems, and most of them use water. 

They also consume vast amounts of electricity – and the production of electricity requires water. 

All of this adds up. British researchers released a report on global water usage earlier this year. That report reveals that data centers use as much as 560 billion liters per year. 

Some forecasts suggest usage will reach 1.2 trillion liters per year by 2030 – meaning it could double within five years. 

The biggest facilities are known as hyperscale centers. They use as much electricity as a small city. They also draw on 2.5 billion liters of water each year – roughly the same amount as 80,000 people. 

In theory, there's nothing wrong with that. But in practice, it may be a different story. 

The water warning

The Global Commission on the Economics of Water (GCEW) issued a warning in 2023, saying the world faced a water crunch. The commission predicted that demand for freshwater will exceed supply by 40 percent at the end of the decade – that's unless conservation methods improve and consumption patterns change. 

Consumption patterns are changing, but often, for the worse. Data centers look set to put more pressure on water supplies, and that pressure seems to be mounting every year. 

Google, for example, used about 13 billion liters of water in 2019. Just four years later, it consumed 24 billion liters. 

There's no doubt that AI has massive potential. Goldman Sachs estimated that it could add around $7 trillion to the global economy over 10 years – mostly through improvements in productivity and new business opportunities. 

But the cost of a water crisis cannot be ignored. The GCEW estimates that by 2050, water scarcity could reduce the value of some economies by eight percent. In some cases, that will translate to hundreds of billions of dollars.  

The fact is, just about every business needs water – whether it's a restaurant or a mine. It's also crucial for the labor market. 

Activist group Water.org says poor infrastructure forces millions of people to source water themselves. It estimates that the global cost amounts to $260 billion per year, mostly because people could be using the time to look for jobs, or start a business. 

Thirsty data centers are now adding another complication to an already complicated situation. But as we grapple with AI's risks and rewards, we are learning a crucial lesson: water is not just an environmental issue. It is a technological issue, and – increasingly – an economic issue. Because without water, there is no AI.