The Ezra Klein Show: The A.I. Revolt Is Here
Journalist Jasmine Sun discusses her reporting on the data center backlash across the American Midwest, exploring how opposition to AI infrastructure reflects deeper concerns about corporate power, democratic process, and skepticism toward AI companies' promises rather than simple NIMBYism or resource concerns.
Summary
Ezra Klein interviews journalist Jasmine Sun about her recent reporting trip to Wisconsin and Michigan investigating opposition to data center construction. Sun describes the physical reality of data centers—massive, windowless industrial facilities with mechanical humming sounds that take over a minute to drive past on highways—and the coalition of activists opposing them, which includes stay-at-home mothers, retired executives, farmers, and environmental groups.
The conversation identifies three converging layers of concern: procedural problems (particularly the use of NDAs that prevent local officials from disclosing project details, creating information vacuums filled by rumors and social media), direct environmental impacts (water and electricity usage, though Sun notes that modern closed-loop data centers use less water than golf courses), and opposition to AI itself as a technology and industry.
Sun argues this is fundamentally different from typical industrial opposition because there is no compelling pro-faction supporting data centers beyond utility companies and AI companies themselves, both unpopular. Unlike solar farms (which have environmental constituencies) or auto plants (which have worker constituencies), data centers lack a clear benefit narrative that resonates with the public. She found that when she asked local organizers if they use AI, most acknowledged limited utility (drafting emails, making memes) but didn't see it as essential infrastructure justifying massive valuations or community disruption.
The conversation explores how AI populism operates—viewing AI not as a normal technology but as an elite political project imposed without consent. This manifests in data center fights through distrust of corporate promises, concerns about job displacement, and fears of being left with stranded assets if the AI bubble bursts (resonating with Foxconn's failed Wisconsin project). Sun recounts the Saline Township case where developers sued a community that voted against rezoning, forcing a settlement—a visceral example of how concentrated corporate power overrides local democratic choice.
Klein and Sun discuss how AI company executives frame challenges as marketing problems rather than product problems, continuing to acknowledge publicly that AI might displace half of white-collar jobs while simultaneously racing to build capacity. They debate whether this constitutes a genuine "race" dynamic with China, with Klein arguing the race metaphor is a central policy dividing line: if recursive superintelligence with a clear winner is inevitable, speed is justified; if not, slowing down for more democratic deliberation is reasonable.
Sun notes an emerging coalition between AI safety advocates (who believe we're in a race but that winning could mean extinction) and populists (who oppose AI billionaire power regardless of safety arguments). Both groups arrive at similar conclusions about slowing AI development through different reasoning. The conversation addresses counterarguments to moratoriums: they may simply shift data centers to less democratic jurisdictions or authoritarian countries, and compute scarcity could exacerbate inequality if demand exceeds supply.
Sun's China reporting revealed less public backlash there, attributed partly to technological determinism (the belief that progress is inevitable) and partly to Chinese state regulations protecting workers and limiting certain AI applications. She dismisses claims that the data center backlash is a Chinese psyop, finding the opposition quite organic and informed.
Klein emphasizes that Silicon Valley has overestimated how much raw intelligence matters versus organizational, political, and relational friction. Most people don't want AI to prove mathematical theorems or grant immortality; they want cheaper goods, better health, and less drudgery. The conversation concludes that Silicon Valley's 2025 political setbacks (failing to control policy despite wealth and influence) demonstrate that politics and democracy are more powerful than tech leaders realize.
About this episode
The A.I. giants weren’t prepared for this.
Key Insights
- Sun found that opposition to data centers is not primarily about water or electricity usage but rather about AI itself and perceived corporate overreach, with people dismissing environmental arguments as post-hoc justifications for deeper distrust.
- The use of NDAs by AI companies prevents local government transparency, creating information vacuums that social media rumors fill, ultimately inflaming opposition more than preventing it—a mistake some companies like Microsoft have now reversed.
- Data centers lack a compelling pro-faction beyond unpopular tech and utility companies, unlike solar farms (environmental constituency) or auto plants (worker constituency), making it harder to build political support even where economic benefits are real.
- The Saline Township lawsuit by data center developers against a community that democratically rejected rezoning exemplifies how concentrated corporate power can override local democratic choice through legal and financial pressure.
- AI company executives frame the data center backlash as a marketing and messaging problem rather than a product problem, despite themselves acknowledging that AI might displace 50% of white-collar jobs.
- Silicon Valley's assumption that raw intelligence solves most problems breaks down when confronted with political, relational, and organizational friction that characterizes governance and implementation in the real world.
- The race metaphor with China is used as justification for unlimited speed in AI development, but this framing precludes alternative strategies like international negotiations or coordinated slowdowns that might actually reduce existential risks.
- An unexpected coalition has formed between AI safety advocates (who believe we're in a race but racing toward existential risk) and populist politicians (who oppose billionaire power), despite fundamentally different reasoning about why AI should be slowed.
- Data center moratoriums in democratic countries risk shifting development to authoritarian nations with less democratic oversight, potentially concentrating AI power in less accountable hands.
- Chinese workers and citizens exhibit less AI backlash partly because of state-enforced technological determinism and partly because the Chinese government has implemented worker protections and AI regulations that the US government lacks.
- People interviewed by Sun acknowledged some AI utility in their lives but characterized it as a toy rather than essential infrastructure, arguing the massive valuations and disruption cannot be justified by current demonstrated benefits.
- Silicon Valley's political influence declined substantially between early 2025 and late 2026, with tech executives discovering that political power cannot be simply purchased despite wealth, and that relational and democratic processes are more powerful than they assumed.
Topics
Transcript
So we are taking this week off along with our producers to get ready for our finale extravaganza, our last ever episode of Hard Fork, which will be next week. And when you see the choreography, you'll understand why we needed a week to prepare. Exactly. There's pyro, there's choreo. It's going to be a big production. There's Oreos. But until then, we have a very special treat to share with you. You may remember past Hard Fork guest Jasmine Sun. She recently appeared on the Ezra Klein Show to talk about her reporting on data centers. Yeah, she took this really fascinating reporting trip through the Midwest this summer, interviewing people on all sides of the debate throughout Wisconsin…
Full transcript available for MurmurCast members
Sign Up to Access