If you have ever wondered why your digital life feels so seamless, the answer lies in massive warehouses filled with humming servers. AI is very thirsty, and in a single year, Google data centres consumed a staggering 6.1 billion gallons of fresh water, according to the Google 2024 Environmental Report.
In the report, the tech giant admitted to the scale of this resource drain. “In 2023, our data centers consumed 6.1 billion gallons of water—17% more water than the previous year, mirroring similar growth in electricity use,” the report said.
This is not just a drop in the ocean; it is enough water to irrigate approximately 41 golf courses annually in the southwestern United States.
As we transition into an era dominated by AI search, it is getting increasingly difficult to quench these machines’ thirst. Behind every prompt is a physical cost that is quickly becoming impossible to ignore.
HOW DOES AI COOLING WORK?
Computers generate intense heat. When you ask an AI to write a poem or generate an image, thousands of processors work simultaneously, creating immense thermal energy.
To prevent these multi-million-pound machines from melting, data centres use cooling towers. These systems work through evaporative cooling, where water is pumped over hot surfaces to absorb heat.
The catch is that roughly 80 per cent of this water literally vanishes into the atmosphere as steam.
Because you cannot recycle steam, it is effectively lost from the local water cycle. This is why a facility can pull in millions of gallons daily but return almost none of it to the local pipes.
WHY IS FRESH WATER NECESSARY FOR DATA CENTRES?
One might wonder why tech giants do not simply use recycled or grey water. The answer is rooted in basic chemistry.
Water naturally contains minerals like calcium and magnesium. Every cycle through a cooling loop causes evaporation, which makes these minerals more concentrated.
Eventually, they form a hard crust called limescale. This scale is incredibly destructive, acting like a thick blanket that traps heat, clogging pumps and eating through expensive heat exchangers.
Recycled water starts with a much higher mineral count, making it even more corrosive from the first second. To use it safely, companies would need to build massive on-site treatment plants.
Currently, less than one per cent of water in the United States is recycled, and most cities lack the infrastructure to deliver it to industrial sites. Consequently, it is cheaper for data centres to simply drink from the municipal tap.
WHAT IS THE ENVIRONMENTAL IMPACT OF AI GROWTH?
The trajectory of this consumption is a major concern for climate scientists. A report by Morgan Stanley suggests that water consumption by AI data centres could increase 11 times by 2028.
Many of these facilities are being built in regions already struggling with water scarcity.
While companies often mention water stewardship projects to replenish what they take, these projects are rarely in the same local watershed where the water is consumed.
It is a system of global offsets that does little to help a small town which is sharing its water main with a facility that consumes as much water as 50,000 people.
As AI becomes more integrated into our lives, the pressure on our most precious natural resource will only intensify.
