The Ezra Klein Show: The A.I. Revolt Is Here

Summary

• Data-center opposition has gone national and bipartisan in under a year, jumping from roughly four-in-ten voters (Heatmap, last August) to seven-in-ten by May, with 100+ local and state moratorium proposals, a one-year New York freeze from Governor Hochul, a DeSantis “citizens bill of rights” for AI, and Bernie Sanders calling for a national moratorium. For anyone underwriting the hyperscale buildout, siting risk is now a first-order cost line — “nobody wants it was the phrase I heard over and over.”
• Jasmine Sun’s core finding from Wisconsin and Michigan: this is not misinformation or classic NIMBYism but a trust collapse in which every pro argument dies on “I don’t believe them.” Marquette pollster Charles Franklin’s mechanism: contested issues poll 50/50; data centers poll 70/30 against regardless of proximity because “there is no strong pro argument” — no constituency beyond utilities and AI companies, “which are already incredibly unpopular.”
• Deal leverage has flipped from developers to towns. Two years ago companies hunted sales-tax exemptions and local subsidies; now officials say “if we did this again today, we wouldn’t have to offer any subsidies,” all the pro-build people Sun spoke to regretted the NDAs, and Microsoft’s Mount Pleasant site is on track to pay $19.6 million in 2026 property taxes to a village of 28,000.
• Stranded-asset fear is the sharpest local objection, anchored in Foxconn’s 13,000 promised jobs that became roughly 1,000 and Janesville’s $30 million of never-remediated GM contamination. The live warning: xAI’s Memphis Colossus was able to sell its compute capacity to Anthropic, which was short of capacity — but “maybe there’s not an Anthropic to pick up the bill” next time, leaving small towns “left holding the bag.”
• Sun doubts the AGI race has a finish line: no agreed definition, “extremely jagged” model capabilities, and no lab “pulling ahead that decisively” despite Anthropic’s coding lead and the major labs’ recursive-self-improvement strategy. Klein’s frame: whether you buy the race metaphor is the central dividing line in AI policy — if you don’t, “let’s run faster to the bad place is not nearly as compelling an argument.”
• Moratoriums won’t, by themselves, slow AI; they’ll relocate compute to Texas, Louisiana, the Dakotas, Australia, Gulf autocracies, or space — while constricting supply so only deep-pocketed buyers get frontier access. With demand exceeding compute and “the closing of the frontier” already underway (the best models “not being open to everybody,” part safety, part pricing), restricting the buildout stratifies who benefits: “maybe my boss is gonna get the superintelligence and automate my job.”
• Watch the strange-bedfellows coalition: DeSantis doing AI roundtables with Max Tegmark, Bernie Sanders doing viral videos with Eliezer Yudkowsky — “I wouldn’t send my kids over to a playdate at the polycule, but I can do coalitions.” Meanwhile China shows little backlash (fatalist determinism plus targeted regulation), and Sun calls the CCP-psyop theory of the US backlash “ridiculous” — the cited tweets have “two views per tweet.”
• Klein’s verdict on the industry’s self-diagnosis: “You don’t have a marketing problem. You have a problematic technology… a product problem.” If job losses land before cancer cures, “politically, that’s not gonna be an equilibrium you can protect” — and Sun sees Silicon Valley belatedly discovering that democracy has friction: “it’s actually really hard to buy an election.”

Deep dive

1. The revolt, quantified — and the view from the fence line

• Klein’s setup: data centers have united Republicans and Democrats when nothing else could — opposition grew from about four-in-ten voters last August (Heatmap) to seven-in-ten by May; over 100 local or statewide moratorium proposals; Hochul’s one-year New York freeze over grid strain and ratepayer costs; Sanders wants a national moratorium “to give democracy a chance to catch up.”
• What a hyperscale site actually looks like: “I felt like I was in Eden… you just see this verdant landscape give way to what is a windowless industrial park. Data centers are very ugly.” Locals time drive-bys at seventy miles per hour — Port Washington’s runs about a minute forty-two. The sound is humming, buzzing, rattling — “inhuman.”
• The organizers: stay-at-home moms, retired executives, farmers, professional environmental activists — “a lot of women, relatively more left-leaning, though definitely some right-leaning folks as well.”

2. NDAs poisoned the process before the concrete was poured

• The mechanism: councils sign nondisclosure agreements and can’t reveal it’s a data center, who the customer is, or the power draw — but whispers escape anyway. A construction-union VP’s Irish saying: “the only way to keep a secret between three people is to kill two of them.” The refrain from local officials: “we could not get ahead of social media because we had signed an NDA.”
• Why do companies demand them? Sun’s blunt answer: “I don’t think there’s a good reason” — they “just thought it would be easier in case they changed their mind.” Microsoft has stopped using NDAs; other labs’ compute teams are considering the same. All the pro-build people she interviewed — workers and AI developers alike — “regret the NDAs.”

3. Water is a sticky icon; electricity is the real bill

• Sun’s correction on water: today’s new builds are almost all closed-loop, recycling water like air conditioning and using “a fraction of the water that golf courses use” — there’s literally a golf course beside the Janesville site. Yet it persists as a symbol: “They’re building right by the Great Lakes. Why would they do that if they weren’t trying to drain the lakes?”
• Electricity “is real”: chips and low-latency interconnects demand significant new generation — “probably natural gas in the near future” — even though an operational data center is relatively clean. The fight migrates to the power plants behind it.
• The tell that grievance can precede evidence, from both speakers: a YouTube video uses more water than a ChatGPT query, while relatively few people in that conversation were giving up meat. Klein: “sometimes people don’t like a thing and they’re looking for reasons to justify that dislike.”

4. “I don’t believe them” — the trust collapse that flipped deal leverage

• The pro case is genuinely material, and Sun defends it against her own side: Microsoft on the old Foxconn site is on track for $19.6 million in 2026 property taxes to a village of 28,000, and she finds it “frustrating personally” when organizers call jobs a myth — 500 maybe six-figure, apprenticeship-track jobs lasting two to six years is “enough time to build a family… to buy a home.”
• But every promise met the same wall: water treatment — “I don’t believe them”; rate guarantees — “DTE has raised our electricity rates basically every year since 2022. Why would this be the year they decide not to?”; jobs — “I don’t believe them. They have big PR teams. They can say whatever they want.”
• The balance of power has inverted: two years ago companies shopped for sales-tax exemptions on GPUs and local subsidies; now officials say “if we did this again today, we wouldn’t have to offer any subsidies,” and developers hunt friendly jurisdictions with “a Christmas tree ornament’s worth of community benefits agreements.”

5. Foxconn’s ghost: bubble fear and the stranded-asset question

• The scar tissue is specific: Foxconn promised Mount Pleasant 13,000 manufacturing jobs, took hundreds of millions in infrastructure and subsidy, delivered about 1,000, and pulled out “because the contract wasn’t set up correctly.” Janesville’s century-old GM plant closed in 2008 leaving $30 million of contamination and forever chemicals that made the brownfield unsellable. “Who’s gonna be left holding the bag? That was the question I heard over and over again.”
• The live illustration: Memphis’s Colossus — “xAI never really took off. People were not in fact using Grok as much as Elon thought” — was able to sell its compute capacity to a fast-growing Anthropic that had not built enough data centers. “You could totally imagine a world where xAI decides we are gonna focus on space… maybe there’s not an Anthropic to pick up the bill.”

6. The abundance test: data centers have no pro-faction

• Klein applies his Abundance frame — “what do we need more of and how do we get it?” — and finds the AI companies can’t win the first clause. Solar opposition used identical tactics (Facebook groups, packed town halls, zoning fights) but solar had a constituency; the Janesville GM plant had seven thousand workers and everyone drives a car. Data centers’ only backers are “the utility companies and the AI companies, which are already incredibly unpopular.”
• Charles Franklin of Marquette Law Polls explains the 70/30: contested issues poll 50/50 when both sides have a case; here opposition is bipartisan and independent of proximity — “it’s not just NIMBYism, it’s you don’t want a data center in anyone’s backyard… There is no strong pro argument.”
• To San Francisco’s favorite gotcha — don’t these people use ChatGPT? — yes: “I’ve used it to draft an email or make a meme.” But they see a widget, a toy, not something essential like cars, energy, or housing — nothing that justifies “these gigantic valuations.”

7. AI populism: Saline Township as democracy microcosm

• Sun’s definition, which Klein calls influential: AI viewed “not only as a normal technology but as an elite political project to be resisted.” The signature complaint isn’t uselessness — it’s “why is this being shoved down our throats?” Her strong claim: “even if there was no misinformation about water use, people would be just as angry about the data centers.”
• The Saline saga: the township council voted four-one against rezoning for the Stargate project; developers sued the few-thousand-person town over “exclusionary zoning”; it settled — “fine, give us a few million for the fire department and for some schools.” To residents it felt like “a profound violation of little-d democracy”: a fundamentally asymmetric deal “that will then transform your lived reality into the world that these tech companies have decided for you.”
• Klein sharpens it: executives testify “it should not just be us making these decisions,” then turn “tremendous amounts of financial artillery” on any community that says no. Sun adds the money paradox — “making the numbers bigger and the amounts of money bigger makes people more suspicious, not less”; Abdul El-Sayed’s top hit on Haley Stevens is how much money she is getting from APAC, DTE, and pharma.
• Nor does she exempt Anthropic: it warns of “fifty percent of white-collar jobs lost by 2030” and Dario’s essays raise “an underclass of people of lower intellectual ability” — while it builds the coding, banking, and design agents that it believes could displace jobs. “They don’t really have a good answer, because their business model is fundamentally premised on the disruption that they say they are causing.”

8. The underclass thesis — and a re-skilling race humans may lose

• The Valley’s permanent-underclass scenario, from Sun’s Times piece: AI and robots do any job a person can, workers lose all economic leverage, capital pays machine labor. Her striking interview finding: “I have not heard a single person in the tech industry tell me that they believe that AI is going to decrease inequality” — some say “the floor will get really high.”
• Her own hedge: she doesn’t think an underclass is the likeliest outcome — “AI is actually just really jagged, and human jobs are super complex and super hard to automate,” and most apocalypse-predictors “haven’t actually worked enough real jobs.” But speed worries her: senior software engineers are doing great and love cloud code, while hiring and job postings are down in that sector; “do we think a human software engineer is going to re-skill faster than the next model is gonna get better at software engineering?”
• Klein’s knot: executives privately tell him “they wish all this would slow down,” yet the AI industry is in “an all-out war to make sure we have as little time for adjustment as possible” — which makes them “a pretty unconvincing pro-faction.”

9. Why build the thing you fear: three rationales and a coin flip

• Sun’s taxonomy of insider self-justification: techno-determinism (“superintelligence is going to be built inevitably… I wanna make it happen in the least bad way”); the expected-value gamble on utopia versus extinction; and pure technical fascination. “People have all sorts of self-justifying narratives… it makes sense to me why the public is not particularly sympathetic to any of these.”
• The gamble incarnate: Sam Bankman-Fried’s podcast answer — Sun thinks it was Tyler Cowen’s — flip a 51/49 coin between doubling human welfare and everyone dying, and he flips. Sun says, “that’s a psychopath”; Klein says the person who flips that coin would have to place a very low value on human life. Sun calls it hyperbolic, but says insiders characterize the superintelligence bet as 90-10 or 80-20 — a risk appetite that makes more sense when people have opted into the endeavor, and is “extremely different when you’re talking about the rest of the world.”
• The backyard question — “why is Mark Zuckerberg not building a data center in his backyard?” — lands for both: Northern California is hard to build in, but Klein insists the deeper truth holds: a data center is “a concentrated cost for a diffuse benefit… if you believe in the benefit,” and communities “feel like pawns in some tech billionaire’s game.”

10. A race with no finish line — and the strange coalition against it

• Sun’s dismantling of the race metaphor: nobody agrees what AGI means, so “the race has different finish lines and moving finish lines”; models are “extremely jagged — super good at math and super bad at poker at the same time.” The major labs are chasing recursive self-improvement — “right now folks think that it’s Anthropic, which has the best coding models” — yet comparing latest Anthropic, OpenAI, and Chinese open-weight models, no one is “pulling ahead that decisively… I don’t know that the race has a finish line, and that’s what worries me about it continuing.”
• Klein’s policy dividing line: if you believe in a ticker-tape race to recursive superintelligence, the next one-to-three years are definitionally everything; if AI is just a powerful technology with heavy downsides, then “let’s run faster to the bad place is not nearly as compelling an argument” — and slowing down for democratic voice “is not crazy.”
• Hence the convergence: DeSantis doing AI roundtables with Max Tegmark, Bernie Sanders doing viral videos with Eliezer Yudkowsky, and one person Sun spoke with saying: “I wouldn’t send my kids over to a playdate at the polycule, but I can do coalitions.” Data centers even got liberal women organizers “to talk productively with their Trump-voting neighbors about politics” for the first time in nearly a decade.

11. The case against moratoriums: capital flight and a closing frontier

• Sun’s first objection: moratoriums don’t necessarily slow AI, they relocate it — companies are already flooding into Texas, eyeing Louisiana, the Dakotas, Australia, and eventually space. Klein extends it: squeezed out of democratic jurisdictions, builds flow to authoritarian ones like the Gulf states — “you might end up with less overall democratic control.”
• Second, deals can be done right: only data-center developer Veridian was willing to clean Janesville’s brownfield after a broker failed to sell it for five years, and the cleared Foxconn site with existing infrastructure was, in her opinion, “a net improvement.” Her fix: state-level or federal research into fair deal terms, so it’s not “a town of twelve thousand negotiating with an OpenAI.”
• Klein’s scarcity point: “we do not look to have excess AI supply at the moment,” so constricting compute means Goldman Sachs and JPMorgan get frontier access while small businesses don’t. Sun ties in “the closing of the frontier” — “the very best models, like Mythos from Anthropic, are not being open to everybody,” part safety decision, part pricing. The public’s fear: “maybe my boss is gonna get the superintelligence and they’re gonna automate my job.”
• The geopolitical angle: allies like Australia, Canada, and Europe are treating hosting as leverage — build the data centers in exchange for guaranteed frontier-model access — work Sun attributes to Anton at Carnegie.

12. China’s non-backlash — and the “psyop” theory of America’s

• From her China trip (caveated: a suppressed information environment, no good polling), Sun found little AI fear — but not optimism. “It was something a lot closer to: technology is a force that cannot be stopped,” eerily like Silicon Valley’s determinism. In a one-party state the question isn’t resistance but “how can I use AI to make sure I don’t fall behind” — upskilling with “OpenClaw or whatever” amid brutal white-collar competition.
• Yet the Chinese state regulates where Washington doesn’t: bans on many companion chatbots (fears about relationships, fertility, addiction), court rulings that “AI can do this worker’s job” isn’t sufficient grounds for layoffs, mandatory labeling of AI-generated images — which gives some citizens “a bit of solace.”
• On the growing tech-elite claim (Kevin O’Leary among them) that the backlash is a Chinese psyop: “I think this is ridiculous” — the OpenAI report’s cited tweets “have like no likes… two views per tweet.” The opposition “feels very organic”; activists know closed-loop from open-loop and have “just personally decided, I’m not that interested in having a data center in my community.”

13. “You have a product problem” — and Silicon Valley discovers friction

• Executives now ask Sun: “Can we do better in marketing? I don’t understand why Waymos are so unpopular… Do we need to cut people’s electricity prices in half?” — with the word “bribe” used more than she’s comfortable with. Klein’s rejoinder: “You don’t have a marketing problem. You have a problematic technology… a product problem. People are not gonna want that future.” Sun on Altman: labs marketed only to recruits and investors — “they never realized everyone else could hear them” — and he changed his tune once the goal became “political goodwill and IPO-ing.” Klein’s subtler read: “people can believe things they don’t feel,” which is why they act heedlessly.
• The values gap: tech utopia is UBI, immortality, and discovering all of math and physics — and “if you do the polling on UBI and immortality, neither are especially popular.” People want things cheaper, health, less crappy work — not “Terence Tao in your pocket” — and it’s “literally technically harder to cure cancer than to prove math theorems.” The recurring suspicion: “Is Peter Thiel just gonna buy himself immortality?” If losses arrive before gains, Klein warns, “politically, that’s not gonna be an equilibrium you can protect.”
• Both attack the intelligence-is-the-bottleneck fallacy. Klein: “the world is full of friction” — drug discovery still needs monkeys, rats, and safety data. Sun: AI researchers’ careers were IC jobs where context lives in a code base; ask how AI solves climate or robotics and you get “I don’t know, it’s just gonna do it” — “a deus ex machina… I find it lazy, actually.”
• Klein’s closing frame: Silicon Valley’s worldview came from “seeing impossible problems prove possible to solve,” DC’s from “seeing possible problems prove impossible” — and now the people who invented AI “are finding it’s impossible to build a data center.” Sun’s 2026 update: the January 2025 DOGE triumphalism has receded; Anthropic’s White House troubles are “largely relational,” not logical — “it’s actually really hard to buy an election.” Her book picks: Labatut’s The Maniac, Carl Benedikt Frey’s The Technology Trap, Priya Parker’s The Art of Gathering.