A Quality Technology Services data center in Fayette County, Georgia pulled close to 30 million gallons of water through two industrial-scale hookups that nobody was billing for, an investigation by county utility officials found. QTS eventually paid about $150,000 in back charges, but the county declined to fine the company, and the overuse coincided with drought notices telling residents to cut personal water consumption. The episode, first surfaced through a public records request and reported on May 11, 2026, has become a reference point for communities approving large AI data center builds without first upgrading their water monitoring.
One of the two QTS hookups had been installed without the utility's knowledge. The other was connected but never tied to a billable account. Fayette County is in the middle of transitioning from older meters to a cloud-based smart system meant to catch exactly this kind of drain, and the water director told Politico that the single inspector covering meters in the county is "spread pretty thin."
Vanessa Tigert, Fayette County's water system director, characterized the lapse as a procedural mix-up and said the county did not want to penalize its largest customer. "They're our largest customer, and we have to be partners," Tigert said. "It's called customer service." QTS was retroactively charged at a higher construction rate but paid no penalty for exceeding the peak usage limits set during the planning process.
Key facts
- 01A Quality Technology Services data center in Fayette County, Georgia drew close to 30 million gallons of water through two unmonitored industrial hookups.
- 02QTS paid roughly $150,000 in back charges at a construction rate, but the county declined to issue fines.
- 03Xylem forecasts AI-associated water use will more than double over the next 25 years, with 40% of data centers in water-stressed areas.
- 04Phoenix-area data centers use about 385 million gallons annually for direct cooling today; Ceres projects that to reach 3.7 billion gallons.
- 05In Utah's Box Elder County, about 4,000 residents paid $15 each to block a 1,900 acre-feet water transfer to a hyperscale site.
QTS pushed back on the framing. The company told Ars Technica it was "false and inaccurate to suggest the facility used any water improperly," adding that "once this billing issue was flagged, QTS paid all charges. All water usage followed relevant and applicable regulations." QTS also noted that county officials disputed resident claims that the overuse caused lower pressure on home wells, since the data center does not draw from wells or groundwater.
“The QTS facility ran two industrial-scale water hookups outside the billing system, pulling close to 30 million gallons before Fayette County caught the gap and recovered $150,000.”— Jaeden Schafer
Construction at the QTS site is expected to continue for up to five more years, and construction draws far more water than the eventual operating phase. The facility plans to run a closed-loop cooling system once online, which does not consume water for cooling during normal operation. That distinction matters for the long-run footprint, but it does not address the construction-phase peaks that are straining the county now.
James Clifton, the attorney and property rights advocate whose records request brought the case to light, said residents are bearing the cost of a partnership tilted toward the largest ratepayer. "The first thing they do is lean on the individuals and the citizens to stop water consumption when we have QTS that's just absolutely draining us," Clifton said. The county insists residents' pressure complaints stem from private wells unrelated to the municipal system QTS uses.
The Georgia case lands inside a broader buildout. Xylem, a water technology company, estimates that 40% of data centers and 29% of global chip fabs sit in water-stressed areas, and projects AI-associated water use will more than double over the next 25 years. The Information reported that the heavier water draw often sits upstream of the data center itself, at the semiconductor fabs and power plants supplying the compute.
Phoenix is the clearest stress test. Ceres estimates data centers around Phoenix already use roughly 385 million gallons of water per year for direct cooling, with that figure rising to 3.7 billion gallons once the region's planned sites come online. A letter to Congress signed by more than 120 organizations warned that data centers consuming up to 5 million gallons a day in drought-prone regions strain drinking water, agriculture, and ecosystems, and asked lawmakers to reject any bills that would fast-track permitting.
Local pushback is producing results in some places. In Utah's Box Elder County, a hyperscale developer withdrew an application to transfer 1,900 acre-feet of water from a ranch to its site after about 4,000 residents paid roughly $15 each to file notices of opposition, the Salt Lake Tribune reported. The Environmental Protection Agency announced a program in March 2026 to strengthen water systems, with early focus on rural areas where utility budgets are thin.
There is a counter-current on the technology side. Microsoft is paying to deploy FIDO Tech's AI-driven leak detection systems, which use sensor data to isolate leaks faster — relevant because roughly 30% of the world's treated water is lost to leaks and theft. The American Water Works Association's 2026 State of the Water Industry report struck a more cautious note, saying utilities are "cautiously exploring new technologies like artificial intelligence, recognizing both their potential benefits and associated risks, especially in the area of cybersecurity."
Skeptics point out that closed-loop systems, while not consuming water for cooling, require treatment chemicals that can pollute waterways if mishandled, and that the electricity powering AI compute carries its own embedded water cost at the generating plant. Consumer Reports reported in March that generating the electricity to run data centers can require more water than the cooling itself. Neither QTS nor Fayette County has disclosed the operational water profile expected after the five-year construction window closes.
The Fayette County story matters less for the dollar figure than for what it reveals about the metering gap between AI infrastructure spend and the municipal systems hosting it. Hyperscale builds are landing faster than counties can upgrade meters, staff inspectors, or write enforceable peak-use limits, and when a billing miss does surface, the customer is too large to penalize. Until water utilities catch up on instrumentation — the kind FIDO Tech and Xylem are selling — the AI buildout will keep producing avoidable disputes that hand ammunition to the organizations now lobbying Congress to slow permitting.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
