The infrastructure powering your ChatGPT habit is now a voting issue, and both parties are scrambling to figure out which side of the NIMBY fence wins elections.
The Summary
- Over half of Americans used AI chatbots in July, yet 70% oppose building AI data centers in their communities, creating a political wedge issue ahead of the midterms.
- A dozen North Carolina towns have implemented moratoriums, two have permanent bans, and New York became the first state to approve a statewide moratorium on AI data centers.
- National labor unions representing tradespeople are now siding with developers, splitting the labor movement along AI infrastructure lines.
- DigitalBridge CEO Marc Ganzi argues power, not chips or software, is the real AI bottleneck, warning against politicizing the buildout.
The Signal
America has discovered the AI version of wanting cheap gas but hating oil refineries. More than half the country is using AI chatbots multiple times a week, but when developers show up to build the warehouses full of GPUs that make those chatbots work, communities revolt. The gap between AI consumption and infrastructure acceptance is now defining local elections and forcing national politicians to pick sides.
The backlash is concrete and spreading. New York passed the first state-level AI data center moratorium in July. North Carolina towns are leading the charge at the local level, with moratoriums popping up like anti-vaping ordinances in 2019. Even Texas Governor Greg Abbott, who initially championed data centers, reversed course and tightened restrictions after resident backlash.
"They're ugly. They're noisy. They sprawl across the landscape. They gobble up energy and water."
The complaints are visceral: noise pollution, grid strain, water consumption, and a general distrust of Big Tech. On Capitol Hill, AOC and Bernie Sanders are calling for a nationwide moratorium, turning data centers into a progressive litmus test. Meanwhile, Trump is caught between his pro-data center rhetoric and a base increasingly hostile to the facilities landing in their counties.
But here's where it gets interesting. The labor movement is splitting. National unions representing electricians, plumbers, and construction workers are siding with developers, opposing bans and moratoriums. The reason is simple: AI data centers need tradespeople, and those workers are commanding premium wages. This fractures the traditional coalition. Progressive politicians want to stop the buildout on environmental grounds. Labor unions want the jobs. Someone has to lose.
Key tensions emerging:
- Environmental impact vs. high-wage blue-collar jobs
- Local control vs. national AI competitiveness
- Energy grid capacity vs. AI model demand
- Big Tech distrust vs. consumer AI adoption
Marc Ganzi, CEO of DigitalBridge, frames the issue as a power problem, not a political one. He's right that the bottleneck is energy infrastructure, but he's naive if he thinks this stays apolitical. The AI buildout is already politicized. The question is which party figures out how to message it first.
The irony is that Americans are voting with their queries. ChatGPT usage is surging. Claude, Gemini, and every other model people interact with needs somewhere to run. Those models don't run on vibes. They run on racks of Nvidia H100s sitting in windowless buildings that need megawatts of power and millions of gallons of cooling water. The disconnect is that people think AI happens in the cloud. It happens in Loudoun County, Virginia.
The Implication
Watch how midterm candidates in swing districts handle this. Data centers are the new fracking debate: jobs vs. environment, progress vs. preservation, national interest vs. local autonomy. If you're betting on AI infrastructure buildout continuing at current pace, you're betting that jobs beat aesthetics. That's historically a safe bet, but the 70% opposition number is loud.
For anyone in the AI supply chain, this is a procurement risk. Moratoriums slow buildout. Slower buildout means constrained compute. Constrained compute means higher inference costs and slower model improvements. The people protesting data centers are the same people asking ChatGPT to write their emails. That tension resolves one of two ways: people accept the infrastructure, or they accept worse AI. Right now, they think they can have both.