Data Centers' Hidden Water Crisis: AI Boom Threatens Local Communities as Tech Giants Avoid Disclosure
The AI revolution is using up water at rates that could rival those of whole cities, but most tech companies won't say how much. As more and more data centers are built in the US to meet the growing demand for AI, people in places like Georgia and Arizona are finding that their wells are running dry. Companies say they can't be open about how much water they use because of privacy concerns.
An internal document from 2022 that was leaked shows that Amazon, the world's biggest data center operator, thought its buildings would use 7.7 billion gallons of water a year by 2030. That's about the same amount of water used by 100,000 American homes. That number only counts "primary" use, which is the water that is directly used in cooling systems. The real effect is two or three times bigger when you count "secondary" water use, which is the water needed to make the electricity that powers these buildings.
The document, called "AWS Water Positive Public Launch," shows executives warning that openness was "a one-way door" and telling people to keep things secret even though they were afraid of being accused of hiding something. One possible negative headline the authors thought of and tried to avoid was "Amazon hides its water use."
This pattern of selective disclosure is common in the whole industry. Amazon has never made public the total amount of water it uses, even though it runs hundreds of facilities around the world. Microsoft and Google, on the other hand, regularly publish data on their water use. The difference makes people wonder about accountability across the board, especially since AI investments are speeding up the building of data centers in some of the world's driest areas.
the scale most people never see
Data centers represent critical infrastructure for modern digital life—cloud storage, streaming services, email, e-commerce, and increasingly, artificial intelligence all depend on massive server farms running 24/7. But most users still can't see the physical reality of digital services. Every ChatGPT question, AI-generated image, and cloud-based computation makes heat that needs to be cooled. Most data centers cool by letting water evaporate.
According to a 2024 report from the Lawrence Berkeley National Laboratory, US data centers used 17 billion gallons of water directly for cooling in 2023. As more people use AI, that number could double or even quadruple by 2028. In 2023 alone, the indirect water use—used to make electricity for buildings—rose by another 211 billion gallons.
To put things in perspective, a medium-sized 100-megawatt data center uses about 2 million liters of water every day, which is the same amount as 6,500 homes. In 2024, one facility in Iowa used 1 billion gallons of water, which is enough to meet the needs of all of Iowa's homes for five days. These facilities are not outliers. As companies try to build up their AI capabilities, they are becoming more common.
According to the most recent measurements, the International Energy Agency says that data centers around the world used 560 billion liters of water each year. By 2030, that number is expected to rise to 1.2 trillion liters. When you use water at home, only 10% of it evaporates. But when you cool a data center, about 80% of the water you use evaporates, making it mostly unrecoverable for local communities.
where transparency disappears
The leaked Amazon document shows how the company decides what to share and what to keep secret. Company officials talked about whether or not to include secondary water use in their accounting when they were planning their "Water Positive" campaign, which aims to return more water than it uses by 2030.
The document says that adding water use related to electricity "would double the size and budget" of the campaign "without addressing meaningful operational, regulatory, or reputational risks." The authors said that customers and the media weren't paying attention to the water used to make electricity, so Amazon only counted direct consumption.
Shaolei Ren, an associate professor of electrical and computer engineering at the University of California, Riverside, said, "In environmental science, it is standard practice to include both to more accurately capture the true water cost of data centers." The selective accounting gives false impressions of how the environment is actually affected.
The authors of the document also said not to share information about Amazon's overall corporate water use, which was about 105 billion gallons per year as of 2021, or the same amount as a city bigger than Houston, Texas. They specifically warned about the "reputational risk of publicly committing to a goal for only a portion of Amazon's direct water footprint."
Even within the Water Positive framework, most of the planned spending was either for projects that would help Amazon's operations by making water more available or for meeting regulatory requirements, not for purely environmental reasons. The document says that about half of the $109 million set aside for "water replenishment" projects would have been spent anyway.
Tyler Farrow, who is in charge of standards at the Alliance for Water Stewardship, said, "It's not right to say that your operations are water positive or water neutral." "No matter what kind of offsetting or replenishment you do, it doesn't mean that the water footprints of your own operations go away."
the local impact nobody discusses
Statistics for the whole country don't show what happens in specific areas. A Bloomberg report found that about two-thirds of the new US data centers that have been built or are being built since 2022 are in areas with high water stress, where demand for water is already close to or greater than supply.
After building a data center in rural Georgia, people say their water supplies have been cut off. Wells that have always provided water for generations suddenly stop working or only work sometimes. Local aquifers are underground water reserves that have built up over thousands of years. They run out faster than natural recharge can fill them back up.
These communities are in impossible situations. Data centers will bring jobs, tax money, and investment in infrastructure, which will help the economy grow. But they also use a lot of water, which is hard for local utilities to handle, especially during droughts or heat waves when both residential demand and the need to cool data centers are at their highest.
The difference is clear: businesses pay water bills at commercial rates, but those rates don't always reflect the true cost of scarcity or take into account the fact that nonrenewable groundwater is running out. Cities that want to grow their economies give out permits without fully checking on long-term water availability. When problems arise—usually years after facilities start running—communities don't have much power to demand changes.
why companies avoid disclosure
The leaked document provides rare insight into corporate reasoning around environmental transparency. Amazon executives identified several concerns about releasing water consumption data:
Competitive disadvantage: Revealing consumption might expose operational details competitors could exploit or enable unfavorable comparisons.
Target vulnerability: Public commitments create accountability. If actual consumption exceeds projections or targets aren't met, companies face reputational damage.
Scope creep: Disclosing water data could create pressure to address broader environmental metrics—carbon emissions, waste, land use, and supply chain impacts—that companies prefer not to quantify publicly.
Regulatory attention: Transparency potentially invites new regulations mandating disclosure or imposing consumption limits currently avoided through opacity.
The document's warning that transparency was "a one-way door" captures corporate anxiety: once you start disclosing, you can't easily stop without appearing to hide deteriorating performance.
Yet this logic assumes public tolerance for secrecy around environmental impacts from critical infrastructure. As water scarcity intensifies and communities experience tangible consequences, that assumption appears increasingly tenuous.
the broader accountability gap
Data center water consumption exemplifies larger challenges in environmental accountability for digital technology. Most people can't see or touch the infrastructure that makes modern digital life possible, like servers, cables, cooling systems, and power supplies.
This lack of visibility makes it possible to separate use from effect. People don't know how much water is used to train AI models or how much electricity is needed to store data in the cloud. The costs to the environment are passed on to the communities that host infrastructure, while the benefits are spread around the world to users and shareholders.
Companies don't have much incentive to become more efficient or use less if they don't have to report their results. The competitive dynamic actually discourages openness: companies that share data about their consumption risk looking worse than competitors that don't share anything, even if their actual performance is better.
Some places are starting to require disclosure. Virginia, which has a lot of data centers, recently made it mandatory for companies to report how much water they use. Other states are also thinking about making similar rules. But there are still no federal standards, so companies can avoid being held accountable in most places.
looking forward
It's clear that the future of AI will continue to drive the building of data centers. AI training and deployment need a lot more computing power than traditional cloud computing. This means that facilities need to be able to cool down more and use more water.
Tech companies defend what they do by talking about how they make things more efficient and how they put money into projects to replenish water. According to Amazon spokesperson Margaret Callahan, AWS has cut its water use per kilowatt by 40% since 2021 and is "53% of the way towards water positive" goals.
But these claims can't be properly evaluated without full disclosure, which includes secondary water use and independent verification. The selective accounting that leaves out water use for electricity, the focus on making things more efficient while overall consumption goes up, and the reliance on offset projects that may not actually add anything all point to public relations priorities instead of real environmental responsibility.
Nathan Wangusi, a former water sustainability manager at Amazon who thinks he was pushed out for speaking out against the company's methods, said, "You don't need to hide or confuse." "It doesn't help you make more money. It makes people less likely to trust you."
When people in areas with little water have to choose between economic growth and water security, it's almost impossible to make an informed choice because there isn't clear information about how much data centers use. People can't weigh their options if they don't know what they are. Local governments can't plan for needs they can't see.
The digital economy will keep growing. The question is whether it grows in a clear way, with honest accounting of environmental costs and real efforts to reduce impact, or whether it keeps hiding consumption and passing on the costs to communities that can't fight back.
0 comments