The public hates data centers, but neither party can figure out how to weaponize that hate.
The Summary
- New NYT/Siena poll shows 61% of likely voters oppose data center construction for AI, with only 14% strongly supporting
- Split is bipartisan: Trump voters barely favor them (49% to 45%), Harris voters lean against them
- This matters because billions in AI infrastructure spending is running headfirst into local opposition, and politicians can't pick a side
The Signal
Data centers are the new fracking. Everyone wants the technology, nobody wants the plant in their backyard, and the economic benefits are too abstract to win over communities staring at warehouses the size of football fields sucking up their power grid.
The poll numbers tell a story about infrastructure in the agent economy that nobody in Silicon Valley wants to hear. Sixty-one percent opposition isn't a marketing problem you ad-spend your way out of. It's a structural conflict between where AI compute needs to live and where people are willing to let it live.
"The split was about even among Trump voters, while Harris voters leaned against — this is a rare issue with no partisan edge."
What makes this dangerous for the AI buildout: there's no political coalition to push it through. With fracking, you had energy independence and jobs. With data centers, you have... chatbots? Agent assistants? The value proposition is too distant, too abstract, too "trust us, this will change everything eventually."
Here's what the opposition sees up close:
- Power consumption that can strain local grids
- Water usage for cooling that competes with agriculture and residential needs
- Property near data centers that becomes harder to develop or sell
- Tax breaks for tech companies while schools need funding
The 14% who strongly support are probably split between true believers in the AI future and people who work in the industry. Everyone else sees risk without reward.
The timing matters too. This poll hit before the recent AI safety slowdown conversations gained steam. If public sentiment on AI capabilities is souring alongside infrastructure opposition, you've got a pincer movement that could slow the entire buildout.
The Implication
Watch for data center developers to shift strategy from "build where power is cheap" to "build where we can win local politics." That means more concentration in states and counties that will trade long-term grid strain for short-term tax revenue. It also means the geography of AI compute could get weird, dictated more by permitting than by optimal infrastructure.
For anyone building in the agent space, this is your supply chain risk. The compute you're counting on might get blocked by a few hundred angry voters in a county planning meeting. Plan accordingly.