AI data centers use far more water than most tech giants report & more related News Here

AI data centers use far more water than most tech giants report

 & more related News Here

Microsoft, Google and Amazon are among the tech companies spending an estimated $1 trillion on AI infrastructure this year and last. In some areas, they are using far more water than reported, depending on how data centers operate. And their water consumption is projected to increase rapidly in the coming years.

These companies produce annual sustainability reports that include water use in their data centers. (pexel)
These companies produce annual sustainability reports that include water use in their data centers. (pexel)

These companies produce annual sustainability reports that include water use in their data centers. But among this group of titans, apart from water used on site, the only water Meta Tally uses is at the power stations that provide them with electricity.

No laws oblige these companies to report the full scope of their water use, both direct and indirect. In the US, indirect water consumption for data centers has historically been about 12 times the amount they consume directly, according to a 2024 analysis from Lawrence Berkeley National Laboratory.

Experts who are calculating the water demand for the largest infrastructure build in US history say it could lead to regional fights over who gets the increasingly scarce resource.

report card

Google’s recently released 2025 sustainability report is an instructive example. The company said it consumed 10.9 billion gallons of water — a 34% increase from 2024 — almost all for data-center cooling.

But how much water was used when generating power for those data centers? Indirect water consumption can vary substantially depending on how the electricity is generated – coal and nuclear power require a lot of water, while natural gas requires less water. There is almost no need for renewable energy, including solar and wind.

According to a paper published earlier this year by Alex de Vries-Gao, a researcher at Netherlands-based university VU Amsterdam, Google indirectly consumes almost three times as much water.

A complicating factor is that Google, like Amazon and Apple, pays for enough renewable energy to equal 100% of the energy the company uses. This doesn’t mean that every electron comes from solar or wind, but it does mean that Google can claim to offset its electricity use with low-carbon-emitting sources that use almost no water.

Some critics – including 16 state attorneys general who signed a letter on 2025 – argue that offsetting energy produced by fossil fuels is not the same as actually replacing those sources of energy. And renewable-energy credits are not the same as offsetting water consumed in a given area. For example, if a river in Nevada dries up, the abundant water in Michigan may not help.

Meta’s indirect water use in 2024 was 19 billion gallons – 20 times greater than its direct water use. META plans to “become water positive in 2030” through water-restoration projects. The company has no plans to reduce indirect water consumption.

Cooling systems inside data centers circulate water to ensure temperature control.
Cooling systems inside data centers circulate water to ensure temperature control.

Similarly, Microsoft has announced data centers that will have zero water use, as well as a commitment to “community-first AI infrastructure.” This includes a pledge to “resource more water than we use”, part of a decades-long effort to reduce water consumption. Like META, it is only responsible for direct water use.

Amazon recently said that its data centers use water seven times more efficiently than the industry average, and the company has reached 75% of its goal of recovering one gallon of water for every one it draws. (This also does not take into account the company’s indirect water use.)

A company spokesperson says Amazon “recognizes the connection between energy and water” and that’s one reason the company has more than 700 wind and solar projects globally.

‘Cheap land and cheap electricity’

Kenneth Gillingham, professor of environmental economics at Yale, says one reason companies have historically not reported indirect water use is that their data centers may be far from power plants. Furthermore, in areas where there is no water scarcity, indirect use of water matters less.

In Homer City, PA, a data-center complex is being built — along with a natural-gas power plant — on the site of a decommissioned coal-fired power plant. Although the site will generate more energy, it is expected to draw the same amount of water as before, including the water needed to cool its data centers, says Jonathan Burgess, director of the Pittsburgh Water Collaboratory, a research and outreach organization at the University of Pittsburgh. He added that it helps that the plant is in an area where water is generally abundant.

Recent analysis by the Guardian and Bloomberg showed that nearly two-thirds of new data-center construction in the US is in water-stressed areas like Phoenix.

A 2025 analysis from Ceres, a nonprofit sustainability advocacy organization, found that today, the total direct and indirect water demand of data centers in Phoenix is ​​about 3% of the city’s annual water use. By 2031, they could exceed 20%, a number close to the total water used by residents to maintain all of Phoenix’s lawns and landscaping.

The Ceres analysis also said water demand could rise sharply in neighboring New Mexico, where some of Phoenix’s energy is produced.

People work at a water recycling plant for SpaceX's Colossus supercomputer in Memphis, Tennessee.
People work at a water recycling plant for SpaceX’s Colossus supercomputer in Memphis, Tennessee.

“Cheap land and cheap electricity have led to data centers being located in high water-stress areas,” says Matthew Pine, chief executive of Xylem, the US water-technology giant that supplies utilities across the country. Additionally, water-hungry, fossil-fueled remnants such as coal-fired power plants are now being kept running long past their intended retirement dates to meet AI demand, he added.

Many new complexes come with their own “behind the meter” natural-gas power plants, some of them huge, such as Microsoft’s future collaboration with Chevron in West Texas. Meta and Amazon also have plans for dedicated, off-grid natural-gas plants for their data centers, and SpaceX and other data-center builders also have plans. This may mean that neighbors are competing less for energy resources, but local water will still have to be shared.

more efficient data centers

Recently, Nvidia said it had solved the data-center water problem, showing off a closed-loop cooling system that doesn’t require additional water after filling. Gillingham, Yale’s professor, says the design appears to be a win on two levels: It reduces direct water use to zero while also reducing the total amount of energy required for cooling.

Microsoft has committed to similar closed-loop technology. In 2024, the company announced that all of its new data centers would use it, starting in 2027.

Unfortunately, most existing data centers use evaporative cooling systems that are energy-efficient but consume water, according to data from the 2024 Lawrence Berkeley Report. Experts say that reinstalling them can be very expensive.

Nvidia's closed-loop cooling system for data centers does not use additional water once filled, sharply reducing water consumption.
Nvidia’s closed-loop cooling system for data centers does not use additional water once filled, sharply reducing water consumption.

Josh Parker, head of sustainability at Nvidia, says, “It makes perfect sense that, as we typically solve water footprint at the data-center operational level, the water footprint for AI potentially becomes driven by power generation.”

However, he says, this should be examined in the context of the amount of energy and associated water consumed by AI data centres. He added, “No other industry in America is subject to the same level of scrutiny as electricity.” He also suggests that AI could help other industries reduce water use.

In these boom times, it is clear why AI data centers are in the spotlight. And the lack of transparency and widespread use of NDAs by many data-center builders has only given rise to suspicion and mistrust. According to climate consulting firm Carbon Direct, these are among the reasons why $170 billion of AI data-center capacity is blocked, halted or canceled by 2024.

“In many cases, whenever a project like this is launched, the information you get is an extremely small amount,” says De Vries-Gao. “You’re just getting the tip of the iceberg.”

Write to Christopher Mims at christopher.mims@wsj.com

Leave a Reply

Your email address will not be published. Required fields are marked *