AINEWS
The first national protest wave against AI data centers lands in a country where some of those facilities are reshaping local water budgets. The math is contested. The disclosure gap is not.
Last week, nearly 150 protests across 42 states made AI data centers a national civic story, the first coordinated national demonstration against them, according to a Gizmodo expert roundup. In Fayetteville, Georgia, a QTS data center campus drew 30 million gallons from the local supply before the city ever sent a utility bill, per records reviewed by Atlanta News First. In The Dalles, Oregon, Google facilities now account for about 40 percent of city water use, roughly 550 million gallons in 2025, according to municipal records reviewed by Water Watch. The headline numbers are themselves disputed. Closed-loop cooling, indirect water use via the local power mix, and cherry-picked baselines all change the math. Underneath the math fight is a more durable one, about who gets to know what a data center draws, and when.
The Fayetteville case is less about volume than about timing. QTS connected to the city water system, drew, and then disputed the bill. The city records show the campus took 30 million gallons before the utility issued any invoice. QTS has said there was no improper use. Either way, the public learned about the draw because a reporter pulled the records, not because the company or the city announced it. The pattern is the same: a single facility can shift a town's water budget before ratepayers hear a word.
In The Dalles, Oregon, the scale is different. Google's data centers accounted for about 40 percent of the city's total water consumption in 2025, according to municipal records reviewed by Water Watch. Two more data centers are on the way. The company has not disputed the underlying figures, but it has argued that a large share of the draw is evaporated and returned to the watershed, and that closed-loop systems make the headline numbers misleading.
The cheapest way to cool a data center is to evaporate water, and that trade is what the math argument is really about. Closed-loop liquid cooling can sharply cut water use at the cost of more electricity upstream. A facility on a dry-region grid that burns a lot of natural gas may pull more water from a different watershed than the one the city sees. A single headline number like "30 million gallons" is a starting point for the conversation, not the answer to it. Most operators, the Gizmodo roundup notes, also keep real-time, facility-level water data internal.
What is not in dispute is the disclosure lag. The public most often learns what a site draws from utility bills or from freedom-of-information requests, after the fact. Latitude Media has reported on Google's effort to standardize a water-use disclosure framework. That effort is voluntary, and there is no federal requirement that a hyperscaler publish the daily draw of a specific facility.
The math arguments will keep moving, and reasonable people will land in different places on whether the total draw is alarming or manageable. Three questions to a local utility survive any verdict on the gallons. How many data centers are connected to this system, and what is their combined annual draw? What is the source water (surface, groundwater, reclaimed), and is the source already stressed? When a new facility connects, how much notice do ratepayers get, and through what channel? If the utility cannot answer, the gap is local and fixable.
The Dalles is about to get an answer to the notice question. The national protest wave landed as local fights over data center water use are producing paperwork. Some states have moved on disclosure legislation. The math will keep getting argued. The disclosure fight is the one that changes who is in the room.