Georgia30 million gallons drawn before Fayette County noticed.
2 of Quality Technology Services’ hookups weren’t billed; one was installed without the utility’s knowledge, during an active drought warning to residents.
The county imposed no fine, calling it a “procedural mix-up”; Quality Technology Services is its largest customer.
Sources: Ars Technica
WHY DATA CENTERS DRINK
A hyperscale data center burns 30–100 megawatts of electricity, and most of that becomes heat. Evaporative cooling — spraying water across heat exchangers so it evaporates and carries heat away — is the cheapest way to dump it. A single large facility can consume 1–5 million gallons per day, comparable to a town of 10,000 people.
THE METERING GAP
US water utilities are mostly municipal, underfunded, and run on infrastructure designed for residential billing. A single industrial hookup can move more volume in a week than a neighborhood does in a year, but the meter, the audit cadence, and the enforcement staff were sized for the neighborhood.
THE CAPTURE PROBLEM
When a single customer becomes a utility's largest revenue source, the utility's incentive to enforce against it weakens. This is regulatory capture at the municipal scale — the regulator and the regulated entity become financially fused, and 'procedural mix-up' becomes the language of non-enforcement.
WHY THE SOUTHEAST
Northern Virginia saturated first; Georgia, the Carolinas, and Tennessee are the second wave. Cheap power from the TVA and Southern Company, low land prices, mild winters, and lax water permitting beat the Southwest's heat penalty. The catch: the Southeast's aquifers and reservoirs are not sized for industrial-scale draw, and droughts are arriving on a tighter cycle.
THE DROUGHT PARADOX
Residents under drought restrictions face fines for watering lawns. The data center next door faces a 'procedural mix-up.' The asymmetry is not accidental — it reflects which class of water user the legal system was built to police.
THE AI MULTIPLIER
Training a single large language model can consume hundreds of thousands of gallons of fresh water for cooling. As AI workloads displace traditional cloud workloads, per-rack power density is rising 2–3×, and water consumption rises with it. The Georgia incident is a preview of a fight that will repeat in every county that hosts a hyperscaler.