They say it many different ways.
“We’re water-neutral.”
“We recycle.”
“We use closed-loop cooling.”
Here is what’s not said: Most of the water a data center actually consumes never touches the building at all.
When a tech company tells a county board that its new campus will use, say, 2 million gallons a year for cooling, that number is real. But it’s only the direct water. They are talking about the water that runs through the chillers on site.
What that number leaves out is the indirect water. You have to account for the water consumed upstream at the power plant that’s burning coal or gas or splitting atoms to generate the electricity that data center needs, 24 hours a day, 365 days a year.
That indirect water, or the water used to make the electricity, accounts for about 80 percent of a data center’s total water footprint, according to industry analysis from Apstech Advisors.
Other research put it differently but lands in the same place. Indirect consumption runs more than 10 times higher than what happens on-site, according to analysis from the Information Technology and Innovation Foundation.
Another study from International Energy Agency in 2025 estimated that about two-thirds of all water used by data centers is linked to electricity generation, while one-fourth is tied to cooling.
Either way you slice it, the number a company puts in its press release is the smallest slice of the real footprint, and it’s the only slice anyone’s required to disclose.
See it in the reports
I wanted to see this dynamic in action, not just in theory. So, I went and read Amazon’s own sustainability materials on data center water use, published earlier this summer.
Here’s what Amazon says about itself.
Its data centers used 0.12 liters of water for every kilowatt-hour of computing in 2025, a figure the company says is seven times+ more efficient than an industry average of 0.84 L/kWh.
Amazon says it’s three-quarters of the way to a 2030 goal of being “water positive,” returning more water to communities than it draws, and that in 2025 it returned three gallons for every four it used. It points to more than 50 water-replenishment projects and 26 facilities running on reclaimed wastewater instead of drinking water.
I didn’t read one sentence in Amazon’s public materials about the water consumed at the power plants generating the electricity behind those servers, or the 80 percent of the footprint that, by the industry’s own analysts’ estimates, actually matters more.
Amazon’s 2.5-billion-gallon global consumption figure, reported this year, is the company’s first-ever public disclosure of a total water number. It’s not a years-long track record, but rather a first draft.
You can access their full Sustainability Report here.
I looked at Google’s environmental reports, which show water consumption climbing from roughly 6 billion gallons in 2023 to 8.1 billion gallons in 2024 to 10.9 billion gallons in 2025, nearly doubling in two years, even as the company touts efficiency gains.
None of these companies’ headline efficiency metrics, the liters-per-kilowatt-hour numbers that show up in press releases and county board presentations, account for water used to generate the electricity that keeps these buildings running.
That’s the shell game, and it’s not unique to any one company. It’s the industry’s shared blind spot, and it’s the one number nobody’s required to hand over.
It’s also worth noting that these numbers come from voluntary corporate reports not audited filings.
I believe California Governor Gavin Newsom was wrong to veto legislation that would have required data centers in the state to disclose water consumption levels, as independent audits can offer a better understanding of what’s going on.
Why does this matter if the total national impact is small?
Nationally, direct data center water use is a rounding error.
By some estimates it’s around 17 billion gallons a year, or about three-tenths of one percent of the public water supply in this country. If you only look at the national average, you could convince yourself this whole fight is overblown. Some people may take that side.
But averages hide a deeper truth. Water isn’t consumed nationally. It’s consumed locally. It comes from your aquifer, your river, or your municipal treatment plant. In the towns where these campuses actually get built, the local numbers look nothing like the national ones.
Take The Dalles, Oregon, with a population of less than 16,000 people. It’s a town with real history. In the 1840s it was the end of the line for Oregon Trail wagons, the point where settlers had to raft down the Columbia because the cliffs made the route west impassable.
Nearly two centuries after pioneers ran out of road there, the town is now running out of water for a different kind of frontier.
Records released after a public-records fight between Google and The Oregonian-OregonLive show Google’s water consumption there grew roughly fivefold between 2012 and 2025 to about 550 million gallons a year.
One company now accounts for nearly 40 percent of an entire town’s water use. The city expects local water consumption to keep climbing in the next several years, with residents’ water bills projected to nearly double as The Dalles scrambles to upgrade its aging water system.
The city still won’t say how much of that rising demand is from Google. Officials point to confidentiality provisions and a promise they made to the company to keep the specific numbers secret. The public bears the cost of the infrastructure upgrade, but isn’t allowed to know whose thirst is actually driving it.
Oregon’s Gov. Tina Kotek said earlier this year that data centers’ impact on Oregon is “not sustainable” at the current growth rate, citing the industry’s consumption of water, energy, and community resources.
The same fight just played out in Virginia.
Google’s planned data center campus in Botetourt County near Roanoke has three buildings, three substations, more than 900,000 square feet, at least $3 billion in investment.
To supply its water needs, the company signed an agreement with the Western Virginia Water Authority, the utility that provides drinking water to customers in the city of Roanoke and the counties of Roanoke, Franklin and Botetourt, for up to 2 million gallons of water a day at the outset, with room to grow to 8 million gallons a day as the campus expands.
When the water authority handed a copy of that agreement to Virginia Business, the water figures were blacked out. The authority’s executive director said the numbers had been shared voluntarily by Google under a promise of confidentiality.
Henri Gendreau, founder of the independent outlet The Roanoke Rambler, sued in Roanoke Circuit Court, arguing the water figures weren’t proprietary and had to be released under Virginia's Freedom of Information Act.
A circuit court judge agreed and ordered the water authority to release the unredacted agreements. The water authority tried to get the ruling stayed while it appealed; the Virginia Court of Appeals declined, and the real numbers finally went public.
A local group, the Southwest Virginia Data Center Transparency Alliance, put it plainly once the documents came out. They were outraged that 2 to 8 million gallons of drinking water a day may go to a single private company.
After a year of litigation the community finally learned how much of its own drinking-water capacity has been quietly committed to one tenant.
It’s also worth noting what the water authority’s own director said about why the number was sealed in the first place. He signs confidentiality agreements tied to economic development projects a couple of times a year. What?!
Redacting water numbers for data centers has become routine administrative practice at a public utility, which means the Botetourt fight isn’t really about one company or one campus. It’s about how many other agreements, in how many other counties, are sitting behind the same kind of non-disclosure clause right now, waiting for someone to sue.
Here’s what the industry will tell you
The counterargument says something like, “It’s not fair to blame the data center for water used at a power plant miles away, any more than you’d blame your own house for the water your utility uses to keep your lights on.”
In a narrow technical sense, that’s true. The data center didn’t build the power plant. It just demands the electricity that keeps the power plant running flat-out.
But that’s exactly the shell game.
A company can announce record water efficiency at its “water-positive” campus, while its electricity contract is quietly driving up water withdrawals at a coal or gas plant two counties over, and nobody at the ribbon-cutting has to say a word about it, because that water isn’t in the data center’s numbers.
What to ask, if you’re at a zoning hearing or a public comment period:
Not “how much water will the facility use,” but “how much water will be consumed, total, including the electricity it draws”? Ask for direct plus indirect.
What kind of power is contracted to serve this load — grid mix, dedicated gas plant, nuclear — and what’s the water intensity of that specific generation source?
Is there an independent, ongoing water accounting requirement, or is the “water-neutral” claim a one-time press release with no audit behind it?
Is the company’s own total-consumption number a track record, or a first-time disclosure with nothing yet to compare it against?
Water usage legislation is gaining steam this year among state lawmakers.
Here’s one toolkit that states can use to address and mitigate the impacts that data centers have on their water resources. A new report also outlines how state legislatures are regulating the water consumption of data centers.
“AI is accelerating data center growth at a pace that traditional water planning was never designed to handle,” co-author Daniel Anderson, J.D., GWC Water Law Fellow said in a statement. “Whether you are a policymaker trying to protect your community’s supply, a researcher tracking utility trends, or an industry leader looking to mitigate risk, understanding these four legislative pathways is vital to navigating our shared water future.”
I’ve spent my career chasing the difference between what a company says on paper and what’s actually coming out of the tap or happening in your backyard. This fight is the same one wearing a new hat.
The water strain is real…. How about those numbers in the press release?
Want to learn more myths and facts about water use at AI data centers? You can check out this newly launched section of the Brockovich AI Data Center Reporting Map.
Want to hear me talk about other issues associated with AI Data Centers beyond the water? Check out my recent interview with Katie Couric.
Learn More About Closed-Loop Cooling Systems
Closed-loop cooling is often presented as the clean solution to data center water consumption. The reality is more complicated.
What is a closed-loop cooling system?
Traditional data center cooling uses open-circuit cooling towers: water is pumped over hot surfaces, some evaporates (removing heat), and the remainder is recirculated. Because water evaporates, it must constantly be replenished from local supplies.
A closed-loop system keeps the coolant (usually water or a refrigerant) in a sealed circuit. It never evaporates or contacts the outside air directly. Heat is transferred to a secondary system, such as a dry cooler (large fans blowing outside air over tubes) or a heat exchanger, and rejected without water consumption.
Some data centers use a hybrid approach: closed-loop primary cooling supplemented by evaporative cooling only during peak summer heat.
✅ Why Closed-Loop Can Be Good
Dramatically reduces or eliminates water consumption — up to 95 percent less water than conventional cooling towers
No cooling tower blowdown (wastewater discharge)
No chemical water treatment required, reducing chemical runoff risk
Less noise from water tower fans in some configurations
Better for drought-prone regions where water scarcity is a real concern
⚠️ Why Closed-Loop Can Be Problematic
Dry coolers require significantly more electricity to achieve the same cooling effect, especially in hot climates
In extreme heat events, closed-loop systems may be unable to cool adequately, requiring backup evaporative cooling anyway
Higher upfront capital cost, which some operators use to justify choosing conventional cooling instead
Some companies claim “closed-loop” status but still use water in the secondary heat rejection stage
The higher electricity draw of closed-loop cooling can increase carbon emissions depending on the grid mix
What questions to ask about any data center’s cooling claims
Is the system truly closed on both the primary and secondary loops, or does the secondary loop use evaporative cooling?
What is the facility’s water usage effectiveness (WUE) rating — ideally close to 0 for closed-loop systems?
Is supplemental evaporative cooling used during peak summer temperatures? If so, at what frequency and volume?
How does the increased electricity demand of dry cooling compare to the carbon footprint of conventional water-cooled systems on the local grid?
Learn more here:
Data’s Hot Secret: The Closed-Loop Paradox
Data Center Water Cooling, HyperFRAME Research, February 2026
Cooling Water Efficiency Opportunities for Federal Data Centers, U.S. Department of Energy
EESI, “Data Centers and Water Consumption,” June 2025
What’s happening in your town? Know about other places that had to sue to get water usage numbers? Let us know in the comments below.
Over 60% of our country is experiencing drought.
Over 20% of our country is experiencing extreme drought.
Data Centers deserve to die of thirst.
Not our people.
They pushed the data centers through under the radar without presenting the financial and environmental impacts. Premature. With all the brilliant minds we have in America and other countries, an environmentally and sustainable design could have been created. But I guess the power and greed got in the way. We need to shut this effort down until a sustainable and safe plan is made and fully vetted. By an external, objective and professional organization.
No community is ready for this current bullshit.