In Today's Issue: What voters decided, the nine claims checked one by one, the problems that are real and how the industry can address them, and what the backlash means for the AI buildout. Records as of November 6, 2026.

Dear Readers,

In Harwood, North Dakota, a town of about 800 people outside Fargo, voters were asked on November 3 whether large industrial projects such as data centers should face an extra review before the city approves them. About three in four said no, so the campus that the developer Applied Digital is building there for Oracle faces no new hurdle. Of the 41 local data-center measures that Ballotpedia, a nonprofit election encyclopedia, tracked that day, Harwood was the only one where the side seeking limits lost. Everywhere else voters chose the ban, the moratorium or the extra rule, usually with more than 70% (Ballotpedia, accessed 11/06/2026; Bloomberg, 11/04/2026).

I want the United States to build more computing capacity for AI. That is why the case against data centers deserves a fact check rather than a shrug. The frontier labs have committed hundreds of billions of dollars to capacity that has to stand somewhere, on land in a county and connected to a grid whose costs households share. Until now the hard limits on the AI buildout were chips and electricity; local consent has joined them. If that consent is being lost to claims that do not survive the data, the industry should say so with numbers. Where the claims are true, it should address the problems.

Opponents bring a long list. Data centers, they say, raise power bills, drink scarce water, hum day and night, pollute the air, strain the grid, create few jobs, pay too little tax thanks to legal exemptions, eat up rural land and are negotiated in secret. Matt Garman, who runs Amazon Web Services, has called the link to power prices "convenient scapegoating" (Bloomberg, 10/10/2026). So which of these claims hold up against the evidence, and which are myths?

All the best,

Kim Isenberg

Of the 41 local measures on November 3, 28 would have restricted data centers and 13 would have eased them. Voters approved 27 of the restrictions, five of them outright bans, and rejected all 13 measures that favoured development. Across more than 250,000 votes, the limiting side took 72.9%, and the median measure went its way with 77% (Ballotpedia, accessed 11/06/2026; our calculation from Ballotpedia's results). Some votes set general rules such as bans or moratoriums; others decided specific rezonings. Wells, Maine, approved a 180-day moratorium with 82%; in Wilmington, Ohio, four rezonings that would have opened land for data centers each failed with about 80%.

The scoreline overstates how representative these towns are. A question reaches the ballot only where residents collect signatures or a council refers it, as 11 councils did here, and half of these votes took place in Ohio. Polls still point the same way. In March, Gallup found 71% of Americans opposed to an AI data center in their area, against 53% who opposed a local nuclear plant (Gallup, 05/13/2026). For most voters, though, the issue was not decisive: in NBC's exit poll, 23% called AI and data centers crucial to their vote for Congress and another 33% a major factor, the lowest of the five issues NBC asked about (NBC News, 11/04/2026).