That’s the question haunting Beverly and Jeff Morris. The retired couple bought their dream home in rural Newton County, Georgia, for the quiet. But since Meta began building a massive $750 million data center just 1,000 feet from their house, the taps have slowed to a trickle. One bathroom doesn’t work. The washing machine sputtered out.
“I’m scared to drink our own water,” Beverly Morris told the NYT.
The couple has spent $5,000 trying to fix the problem. But replacing their well would cost five times that and they don’t have the money. Meta denies responsibility, but three of the Morrises’ neighbors also report similar water issues. And across the county, a larger crisis is quietly unfolding.
Thirsty Centers
Every time a data center trains a powerful AI model or serves up a social media post, it generates heat. That heat must be removed to keep the computers running, and the cheapest, most common way to do that is to pump in water to cool the machines. This is true for all types of data centers, but especially for AI centers.
The core driver of AI’s water demand is the dramatic increase in the power density of its processing units. Unlike conventional data centers, AI facilities are packed with high-performance Graphics Processing Units (GPUs) and custom accelerators that consume far more electricity and therefore dissipate far more heat within a smaller physical space than traditional Central Processing Units (CPUs). This concentration of energy creates a “thermal wall” that needs a lot of water to cool down. Conventional data centers have often relied on air-based cooling, circulating chilled air through server racks. However, this method is inefficient for the high-density heat loads of AI servers, which need water.
A single older data center like Meta’s in Newton County can use 500,000 gallons of water per day — enough to supply thousands of households. Newer, even thirstier centers are seeking permits for millions of gallons daily. One permit in Georgia requested nine million gallons a day, the equivalent of 30,000 homes.
Choosing Location Above Resources
Maintaining optimal operating temperatures for this sensitive electronic equipment is non-negotiable for companies. Yet these companies prioritize locations with cheap energy rather than large water availability, because energy is more expensive than water. Water is almost regarded as an afterthought. But for communities struggling, it’s becoming a real problem.
To make matters even worse, much of this water is entirely consumed. In evaporative cooling systems, the vast majority of water is lost to the atmosphere and not returned to the local watershed. This makes data centers different from other industrial processes, where a larger portion of the used water is treated and then discharged back into the system.
Researchers estimate that a series of 20 to 50 queries to a large language model like ChatGPT can require the consumption of roughly half a liter (about one bottle) of fresh water for cooling the servers processing the request. An image like the one above can also require around half a liter of water. When multiplied by the billions of such interactions occurring daily worldwide, this “invisible” water cost becomes a substantial driver of the industry’s overall consumption.
The People Caught in the Middle
But that’s just the direct cost. The indirect cost comes from AI centers competing for water with coal, natural gas, and nuclear facilities, which are also highly water-intensive.
In the middle of this, of course, are the people.
Newton County isn’t alone. From Phoenix to Colorado to the UAE, communities are wrestling with water shortages linked to booming data center construction. Around Phoenix, developers have paused homebuilding due to drought exacerbated by tech facilities. In Colorado, water-hungry data centers are now part of renegotiations over the Colorado River.
And the data is murky — literally and figuratively. Companies rarely disclose their water usage, making it hard for regulators to act. Georgia’s legislators tried to push back with a bill to repeal tax incentives for new data centers, but Governor Brian Kemp vetoed it, citing the risk to “economic development.”
This also comes at a time when many cities across the United States and beyond are already grappling with prolonged droughts fueled by climate change. In the Southwest, record-low reservoir levels have triggered emergency water restrictions and climate models predict these dry spells will only become more frequent and intense. Adding the large, largely opaque demands of AI data centers to this already fragile system risks tipping some areas into full-blown water crises, where the basic needs of people must compete with the computational thirst of machines.
Left High and Dry
The Newton County Water Authority is racing to upgrade recycling facilities to keep pace, but even that won’t be enough without drastic action — or cooperation. So far, there’s no sign Meta or other tech giants will help foot the infrastructure bill. A Meta spokesperson said the company conducted its own study on the Morris property, concluding it was “unlikely” that its data center affected their groundwater.
Mike Hopkins, executive director of the Newton County Water and Sewerage Authority in Georgia, said he’s seeing applications requesting up to 6 million gallons of water per day, more than the entire county’s current daily usage. Already, Meta’s data center gobbles up around 10% of the county’s total water use daily.
In the rush to power smarter machines, we may be leaving entire communities drier, poorer, and forgotten. And as Beverly Morris put it, perhaps most haunting of all: “It feels like we’re fighting an unwinnable battle we didn’t sign up for.”
The most significant developments seem to be in the US, but elsewhere in the world, data centers are causing similar distress. Uruguay had its worst drought in history in 2023. Google’s plan to build a data center that would consume 7.6 million liters of potable water per day sparked widespread public outrage. Protestors adopted the slogan, “No es sequia, es saqueo!” — “It’s not drought, it’s pillage” — framing the issue as a corporate appropriation of a vital public resource at a time of crisis.
In the Netherlands, where much stricter water laws exist, a 2021 investigation revealed that a Microsoft data center had consumed 84 million liters of drinking water. The figure was much higher than the 12-20 million liters the company and local officials had publicly estimated.
What Can Be Done?
If we are to improve the situation, policymakers should start treating water as a critical and finite resource, not simply a disposable afterthought.
First, we need transparency about water usage. Without transparent reporting of water, governments can’t make informed decisions even if they wanted to. Then, we need to factor in this water usage and ensure the community’s water reserves can sustainably support it. If a data center brings jobs and investments into an area, a part of that money should go to water investments.
The AI revolution is in full swing, and demand for water will continue to grow. Ultimately, site selection for future data centers should prioritize water sustainability alongside energy costs, ensuring that communities aren’t left dry in the race to power artificial intelligence.
Otherwise, more communities will be left in battles they can’t win.
