The Shocking Truth About Water-Guzzling Hyperscale Data Centers
We are witnessing one of the most breathtaking acts of public deception in modern history, and it centers on the simple claim that artificial intelligence data centers don't consume water. Commerce Secretary Howard Lutnick recently stood before the public and declared that data centers don't use water. This is not just inaccurate; it is a dangerous, deliberate lie designed to obscure the true environmental catastrophe of the AI buildout.
As I detailed in my own coverage of the data center boom, these facilities are not the clean, ethereal 'clouds' the industry wants you to believe -- they are industrial parasites that consume the very resources we need to survive.
What makes Lutnick's claim so infuriating is its sheer audacity. Just a year ago, he was singing a completely different tune, acknowledging that these facilities demand immense volumes of water to cool their machinery. The contradiction isn't a slip of the tongue; it is a calculated pivot designed to fast-track a corporate takeover of our energy grid and water supply.
Lutnick and Trump, among others, are running a top-down misinformation campaign, and those of us who have been sounding the alarm about Big Tech's predatory expansion are now being proven right. The stakes have never been higher, as local communities are forced to sacrifice their most vital resources at the altar of Silicon Valley's delusions of grandeur.
The Direct Impact: Billions of Gallons Vanishing Into Cooling Towers
Let’s move beyond political spin and look at the hard, unassailable numbers that the Commerce Secretary refuses to acknowledge. The water consumption figures released by the tech giants themselves reveal a shocking reality. Google admitted to consuming 7.7 billion gallons in 2024, while Amazon's AWS division guzzled another 2.5 billion gallons. To put that in perspective, a typical data center will use about 300,000 gallons of water per day, but hyperscale facilities -- the ones being rushed into construction across the country -- can drain upwards of 5 million gallons daily (each). This isn't subtle resource management; it is encroaching on limited local water resources, given that roughly 40% of new data centers are being built in high stress or extreme stress water resource regions.
The bulk of this water vanishes into massive open-loop evaporative cooling systems, where it is simply vented off into the atmosphere to dissipate heat. This means the water is permanently lost from the local hydrological cycle, torn away from the aquifers and rivers that communities, farmers, and ecosystems depend on. As I've investigated in the data center wars, this consumption is not merely hypothetical; it is accelerating the transformation of fertile lands into parched deserts.
When water options are being taken away from citizens to supply the whims of server farms, we are crossing a moral boundary that no technological advancement can justify. The water they take today is the water our children will need to drink tomorrow. But it won't be there for them. Entire regions of the USA will collapse from groundwater depletion within one generation.
While the water flowing directly into cooling towers is staggering, it represents only a fraction of the true water consumption incurred by the AI industry. The hidden drain is the water required to generate the colossal amounts of electricity these facilities demand. In 2023, the power plants serving U.S. data centers consumed an estimated 211 billion gallons of water to cool their own operations. That is over ten times the volume of the direct cooling water mentioned earlier -- water that is being burned off in thermoelectric power generation (mostly using hydrocarbons) just to keep the algorithms running.
To ignore this connection is either a display of profound ignorance or a calculated act of fraud. Anyone assessing the environmental footprint of AI must account for the water used by coal, nuclear, and especially natural gas plants (gas turbines) that power these facilities. I reported on a new natural gas-powered AI data center in Pennsylvania, which shows that many of these gas turbines also require water for injection and cooling, further compounding the issue.
When we see that every gigawatt of AI "innovation" relies on the evaporation of precious water, the promise of a "cloud" becomes a tornado of resource extraction, leaving behind dry wells and broken communities in its wake.
Industry defenders often retreat to a familiar talking point: that newer "closed-loop" systems don't actually consume water. But this is a myth designed to obscure a toxic secret. Even in closed-loop systems, water must be periodically purged to manage the "cycles of concentration." As water pulls elements from circulation pipes and fittings, the dissolved minerals and chemicals become more and more concentrated. To keep the system from scaling up and corroding, that deadly brine -- called blowdown -- must be discharged and replaced with fresh water.
This blowdown is not clean water. It is a chemical cocktail laced with corrosion inhibitors, heavy metals, nitrites, and in many cases, PFAS "forever chemicals" that are used to protect the sensitive equipment. These poisons are routinely dumped into municipal sewers, local rivers, and even aquifers. As I have written before regarding data centers and the toxic alliance fueling a public health crisis, there is virtually no federal oversight or requirement for transparency on these discharges. This lack of regulation makes hyperscale data centers a major point-source polluter, free to contaminate the waterways of rural America without consequence.
We are allowing these corporations to turn our public water supplies into private toxic waste dumps, and the silence from the EPA is deafening.