The Myth of Water Usage in Datacenters
How Bad Math Created a Panic
You’ve probably seen the headlines: AI is guzzling water. Data centers are draining wells. Communities are running dry while tech giants train their chatbots. One popular book even claimed a single Google data center in Chile would use a thousand times more water than an entire city.
There’s just one problem: the math is completely wrong.
Consider these common claims:
“Texas data centers are forcing residents to take shorter showers.” The reality? Data centers added about 0.005% to Texas’s water demand—equivalent to 1,600 people moving to a state of 30 million.
“Data centers are built in water-stressed areas.” In Maricopa County, Arizona—one of the most stressed areas—data centers use 0.12% of the county’s water. Golf courses use 3.8%. Per gallon consumed, data centers generate 50 times more tax revenue than golf courses.
Where’s the Actual Impact?
Here’s what investigative reporting hasn’t found: a single instance where data center operations have raised household water bills in America.
Not in Loudoun County, Virginia—home to the world’s largest concentration of data centers.
Not in The Dalles, Oregon, where 29% of city water goes to a Google data center. When rates rose in 2025 for infrastructure upgrades, Google’s fees actually reduced the impact. Without the data center’s revenue, household bills would have increased 23%. Instead, they rose just 7.3%.
Not in Mesa, Goodyear, or any of the dozen other locations where reporters have sounded alarms.
The reason? America separates residential and commercial water markets specifically to prevent industries from competing with households. When data centers require infrastructure upgrades, they typically fund them—like Microsoft’s $31 million contribution to a water recycling facility in Quincy, Washington, that now serves the broader community.
The One Real Story—That Wasn’t About Water Use
The New York Times ran a story about Meta’s data center in Social Circle, Georgia, where nearby wells dried up. Buried in the article: the data center hadn’t been turned on yet. The problem was construction sediment—an issue that could happen with any building project.
The subtitle still read: “Tech giants are building data centers that guzzle up water. That has led to problems for people who live nearby.”
That’s not misleading—it’s false.
Why the Numbers Sound Scary
Journalists love comparing data centers to households, but households account for only about 1% of Americans’ total water footprint. The other 99% goes to producing food, generating electricity, and manufacturing goods.
All data centers combined use about 0.05% of U.S. freshwater—less than the water footprint of Paterson, New Jersey. We spend half our waking hours online. It’s almost miraculous that this infrastructure uses such a tiny fraction of our resources.
Another trick: reporting maximum permitted water use as actual usage. Permits cover worst-case scenarios—every server at peak capacity on the hottest day. Actual usage typically runs 20-50% lower.
What About the Future?
Data centers are growing rapidly. Some estimates suggest water consumption could quadruple by 2028. But even quadrupled, we’re talking about 0.2% of national freshwater consumption—still less than the lifestyle footprint of 800,000 people in a country that adds 1.6 million people annually.
The real question isn’t whether data centers use water—everything does. It’s whether they use it efficiently relative to value created. Data centers generate roughly $20,000 in revenue per thousand gallons consumed. Agriculture generates $19. Electric power generation generates $312.
The Bigger Picture
Local communities should absolutely scrutinize data center projects and negotiate fair terms. There are legitimate environmental concerns about AI—energy consumption, e-waste, carbon footprints from training models. In some cases, like xAI’s facility in Memphis, there are real air pollution concerns.
But when math errors are off by factors of thousands, when construction accidents are misreported as operational failures, when 0.005% impacts are framed as catastrophes—we’ve lost the plot.
The irony? This alarmism makes it harder to address real environmental issues. When people discover the “water crisis” was wildly exaggerated, they become skeptical of all environmental concerns about AI—including legitimate ones.
The AI water crisis is real in one sense: it’s a crisis in how we verify sources and report on technology. Until we fix that, the most wasteful thing about AI might not be its water use—it might be all the attention we’re wasting on a problem that barely exists.
References
[1] https://andymasley.substack.com