← Latest AI News

November 9, 2026 · Daily brief

40 out of 41 towns voted to restrict data centers this week

Sovereignty angle
If towns start blocking the buildings that power ChatGPT and Claude, two things happen: your cloud bill climbs, and the local option — running AI on your own computer — starts looking smarter.

US voters backed data center restrictions in 40 of 41 ballot measures during midterm elections, citing power use and water consumption.

What changed

41 data center ballot measures appeared in US midterm elections on November 3. Voters in 40 towns chose to ban construction, impose moratoriums, or reject projects. The lone exception was Harwood, North Dakota.

Ohio led with 21 measures. California, Florida, Illinois, Kansas, Maine, Nevada, New Jersey and Wisconsin also voted. 77% of voters rejected data center development, according to results compiled Wednesday.

Tech companies outspent local campaigns roughly 100 to 1. Voters chose restrictions anyway. Newly elected Democratic governors and members of Congress in Ohio and Texas ran on data center opposition. A nationwide poll found that more than half of midterm voters called AI and data centers a major factor in their vote.

What you can do with this now

Expect cloud and AI service prices to climb over the next 12 months. If data centers face permitting delays or outright bans, the companies that run them will pass construction costs and scarcity premiums to customers.

Review what you're paying monthly for ChatGPT Plus, Claude Pro, Midjourney, or any AI tool billed per seat or per use. If you're spending more than $100 a month across your team, compare the cost to running a local model on your own hardware.

Tools like Ollama let you run capable open models — Llama, Qwen, Mistral — on a decent laptop or desktop. The output quality lags behind GPT-5 or Claude Opus, but for drafts, summaries, and internal work, it's often enough. And the marginal cost per task is zero.

What to keep in your own hands

Don't move mission-critical workflows to a local model without testing it on real tasks first. Run it alongside your current tool for two weeks. Compare the results. If the local model misses nuance or gives you three bad outputs for every good one, the time you lose fixing errors will cost more than the subscription.

Watch your contract renewal dates. If your AI vendor raises prices and you're locked into an annual plan, you'll pay the higher rate for the full term. Monthly plans cost more per month but give you the option to walk away.

The vote against data centers is about power and water, not about your business. But the supply crunch it signals is real. Budget for it now, and keep one workflow that doesn't depend on the cloud.