theguardian.com web signal

Erin Brockovich turns activism toward AI datacentre boom

TL;DR

  • Brockovich's April callout pulled 3,862 replies in a month from US residents worried about local AI datacentres.
  • Per Guardian analysis, two-thirds of planned US datacentres sit in drought-stricken areas, with larger sites drawing up to 5m gallons of water a day.
  • Her tracker showed 33 AI datacentres operational, 68 under construction and 41 proposed as of 24 June.

Erin Brockovich, the consumer advocate whose fight against Pacific Gas and Electric was made famous by a Julia Roberts film, has spent this spring building a public map of America's AI datacentres and the towns trying to push back. Reporting from the Guardian describes a project that began in April with a callout on her website, and within a month she had 3,862 replies from residents worried about a facility coming to their town.

The scale she is drawing attention to is real. Per the Guardian's own analysis, two-thirds of planned US datacentres sit in drought-stricken areas, and the larger facilities can pull up to 5 million gallons of water a day for cooling, roughly the average usage of 50,000 people. She reads out one resident's water bill that jumped from $22 to more than $350 in a month. As of 24 June, her tracker showed 33 AI datacentres operational, 68 under construction and 41 proposed.

Why this is interesting if you do not live next to one of these sites: the buildout that powers frontier models is happening through utility filings and county-level land deals that often move faster than the people affected can keep up with. Brockovich's frame in the piece is bluntly about asymmetry. She tells the paper that "we have to have some courage to show up, and it's difficult to do that when you're up against forces that have all the money and all the intelligence and all the bandwidth in the world." That is the political problem the AI compute boom has not yet had to solve.

The honest caveat is that this is one campaigner's view, and the Guardian piece is a profile rather than a forensic audit of any single project. It doesn't break out how many of the 3,862 reports describe concrete harm versus general anxiety, and it doesn't separate how much of the water and power load comes specifically from AI training versus other cloud workloads. What it does establish is that the opposition is no longer scattered. She tells the paper she has now been contacted by people in Australia, India, Scotland and Ireland.

What to watch from here is whether hyperscalers and the utilities backing them respond with real disclosure of water draws, grid contracts and ratepayer impacts, before the fight gets organised at her scale. The PG&E case is the precedent she carries, and this time the counterparty is most of the AI industry's capital plan.

Shared on Bluesky by 2 AI experts