New York City Mayor Zohran Mamdani has a phrase for how artificial intelligence gets built in America today: our risk, their reward. In WIRED’s Big Interview podcast, recorded at City Hall ahead of the November midterms, the mayor laid out an unusually blunt civic case against the AI data centers spreading across the country — and a concrete playbook for governing AI data centers themselves where Washington will not.
Sitting down with Katie Drummond almost exactly a year after WIRED first interviewed him as a candidate, Mamdani moved from electricity bills to an AI bubble warning, from banning generative AI in public schools to deepfakes that even his own family cannot spot.
The interview is the clearest statement yet from a major American mayor on AI infrastructure politics, and it lands at a moment when data center opposition has become one of the few genuinely bipartisan causes in national life. This piece unpacks what Mamdani said, what he has actually done in ten months of governing, and what his argument means for the AI build-out economy.
In This Article
- Mamdani’s AI Data Centers Doctrine: ‘Our Risk, Their Reward’
- The Energy Math Behind The AI Data Centers Revolt
- A Bipartisan Backlash Against AI Data Centers
- The Bubble Comparison Mamdani Is Putting On The Record
- New York City’s AI Playbook: Schools, Realtors And Chatbots
- Job Meaning And The Emotional Cost Of AI
- Deepfakes, Democracy And The Crisis Of What Is Real
- What Mamdani Wants Washington To Do
- References
Mamdani's AI Data Centers Doctrine: 'Our Risk, Their Reward'
The phrase itself is the thesis. When Mamdani describes the AI economy as “our risk, their reward,” he is naming a specific transfer: communities absorb the grid strain, the construction disruption and the environmental cost, while the profits and valuation upside flow to a handful of companies and their investors. That framing, from a mayor with a 69 percent approval rating, gives local opposition to AI data centers a political vocabulary it previously lacked.
The context matters. This is Mamdani’s second WIRED sit-down — the first, in October 2025, came weeks before his historic election win and earned him WIRED’s “internet’s Mayor” label. A year on, the electricity in the room has dimmed slightly but the arguments have sharpened. He governs now, which means his AI opinions come attached to things his administration has actually implemented.
His starting point is information asymmetry. Mamdani argues the public debate over AI data centers happens on deliberately incomplete data: opacity around the true energy consumption of AI data centers and the true cost of a query is “by design,” and it exists because powerful companies benefit from the technology being treated as mythical rather than as measurable infrastructure. You cannot price what you cannot see, which is exactly why the numbers around AI data centers stay hidden.
The Phrase That Does The Work
Political language earns its keep by making an invisible transaction visible, and “our risk, their reward” does exactly that for AI data centers. Utilities upgrade transmission for AI data centers at regional cost; the training runs that need them belong to firms headquartered elsewhere. Zoning fights pit one town against one hyperscaler; the valuation model behind the build is global. Mamdani’s phrase collapses all of that into six words a voter can carry to a kitchen table.
What separates this from a politician’s soundbite tour is specificity: a supported state moratorium, city rules with teeth, and a school-system ban aimed squarely at children’s technology exposure. Each of those threads gets its own section below.
| The Setup | Detail |
|---|---|
| Interview | WIRED Big Interview podcast with Katie Drummond |
| Location | Mayor’s office, City Hall, New York |
| Date | Published Nov 3, 2026, ahead of midterms |
| Tenure | About 10 months as mayor; turned 35 in office |
| Standing | 69% approval among New York City voters |
| Headline line | “Our risk, their reward” on AI data centers |
The Energy Math Behind The AI Data Centers Revolt
One Training Run, Three San Franciscos
Mamdani’s energy argument leads with a single yardstick, and it is a good one: training a version of ChatGPT took roughly the same amount of electricity as it would take to power all of San Francisco for three days. That is one model, one training run, one company — and it is consumed by AI data centers rather than spread across households. The question he puts to voters is what a thousand such builds does to a neighborhood’s electricity rate.
He keeps the argument grounded in household ledgers, not abstractions. Opponents of AI data centers, he notes, are objecting to what these facilities mean “in their day-to-day lives, what it would mean in their electricity rates, what it would mean in terms of the impact on their environment.” It is the same cost-of-living lens that powered his mayoralty: rent, childcare, groceries — and now the utility bill that AI data centers help run up.
Our piece on how AI has come for Arizona’s water documented the other half of that ledger — the volume of water that cooling demands pull from arid communities. And our earlier grid analysis showed the counter-proposal the industry favors: our report on data-center flexibility found AI data centers could absorb grid limits with as little as 35 hours of reduced demand, a technical answer that rarely reaches the households paying the bills.
That asymmetry is why Mamdani keeps returning to disclosure. Until AI data centers report their actual power draw, water use and rate impact in terms the public can audit, he argues, every pros-and-cons debate is staged with one side holding the receipts. The energy math is not a footnote to the AI story; in his telling, it is the ballot issue the AI story becomes.
A Bipartisan Backlash Against AI Data Centers
Where The Opposition Is Winning
The most striking thing in Mamdani’s data-center account is who he says is in the movement. He cites Florida Governor Ron DeSantis — politically about as far from a democratic-socialist mayor as American politics allows — as sharing a similar opposition to new AI data centers. When the same complaint animates the Bronx and the Panhandle, the fight has cut across the usual lines that separate Americans.
Mamdani also points to California, where municipalities have begun moving to ban data center construction entirely — not slow-walk it, ban it. Localities from Virginia to the Midwest have passed moratoriums of their own in recent years, and New York state itself is considering the one-year pause on data center construction that Mamdani has publicly supported. A pause, in his telling, is not prohibition; it is the time needed to gather the cost data the industry keeps out of view.
Why now? Two reasons, in Mamdani’s telling. First, AI data centers have stopped being hypothetical renders and started being fence-lines — the fight moved from speculation to bulldozers. Second, the electricity conversation already reached household budgets, so the promise of cheap growth no longer arrives uncomplicated. Local ordinances against AI data centers keep passing because voters can name the cost and the bill, not just the hype. A moratorium becomes attractive precisely because it looks, in a year of fast-moving construction, like the only tool that actually slows time down.
The industry’s physical footprint keeps the issue alive on the map. OpenAI’s pivot to “AI compute campuses” recasts hyperscale builds as university-style developments, complete with promises of landscaping and community amenities. And in Memphis, our report on the xAI data center lawsuit covered a federal judge handing the NAACP a stunning early win over the company’s unpermitted gas turbines — proof that the risk Mamdani describes is not theoretical.
| Place | Action | Status signal |
|---|---|---|
| New York state | One-year pause on data center construction | Mayor supports; state-level push |
| California | Municipalities banning construction outright | Local ordinances advancing |
| Florida | Governor-level opposition to new builds | Bipartisan flashpoint |
| Memphis, TN | NAACP suit over unpermitted xAI turbines | Court win for opponents |
| Nationwide | “Our risk, their reward” framing | Cutting across party lines |
The Bubble Comparison Mamdani Is Putting On The Record
Housing-Bubble Logic, AI Edition
Asked point-blank whether he worries the AI economy is a bubble, Mamdani answered “yes” twice — and then went further than the question. His core comparison is structural: there are “far too many similarities between the housing bubble and the state of AI in our economy today.” He names the mechanism too: business dealings that look like attempts to “prop up a completely artificial valuation” of the companies involved, which in turn justifies the money they spend daily.
The tell, in his account, is the wildly varying estimates of actual profitability at the frontier AI companies — numbers so divergent that no outsider can tell whether the economics work at all. That is the same opacity he objects to on energy, arriving in the financial domain: when only the insiders can measure the thing, the public is asked to trust, not verify. The capital flowing into AI data centers magnifies the problem — these are among the most capital-intensive constructions on earth, underwritten by valuations no outsider can check.
He does not trust. “Do you trust the Sam Altmans of the world?” he shot back at Drummond — before answering his own question in the negative. The parallel to 2008 is uncomfortable for the industry: self-regulation arguments (“this is too complex for outsiders”), valuations sustained by circular financing, and costs that get socialized when the reckoning arrives.
Our reality check on the AI spending boom — the capex year, the free-cash collapse at hyperscalers, the debt loads — reads like the footnotes Mamdani’s argument supplies. Nvidia bulls and bears fight the same question from the other side: brilliant strategy or busted bubble. Mamdani is putting an elected official’s verdict on record while the accountants are still arguing.
New York City's AI Playbook: Schools, Realtors And Chatbots
Schools, Listings And Doom Loops
Mamdani’s answer to federal abdication is a list of things New York has actually done, and it reads like a municipal AI policy first draft for the country. The centerpiece is a proposed moratorium described as one of the most comprehensive in any American school system: a ban on generative AI for students from 2K through eighth grade. His justification is developmental — students need room “to be students,” to try and fail and learn — and pointed: an end to using students “as guinea pigs for the profit motives of technology companies.”
The pharmaceutical analogy is the line to remember. Mamdani notes that nobody would let a drug company help write a science curriculum, yet schools have been “all too willing” to let technology companies shape both the products in classrooms and the teaching itself.
Two enforcement actions show the range of the toolkit. The Department of Consumer and Worker Protection ruled that realtors cannot use AI-generated images of apartments — Mamdani’s verdict on the gorgeous-but-fake listings: “They looked incredible, but they were not real.” And his administration drafted another rule to limit what he calls the “doom loops” between New Yorkers and chatbots. The harms they target are documented — our coverage of the OpenAI lawsuit over GPT-4o and seizure medication, and Disney pressing Character.AI into bankruptcy as, in our telling, a get-out-of-lawsuits card.
These are local instruments against a global technology: realtor rules and chatbot limits stand in for the federal law that does not exist. The municipal-speed logic extends to infrastructure too — Mamdani supports slowing construction of new AI data centers until municipalities can read real energy disclosures, precisely the kind of rule a city or state can write this year while federal AI law remains a blank page.
| Measure | Instrument | Target |
|---|---|---|
| Generative AI ban, 2K–8th grade | Proposed school-system moratorium | Students and classroom practice |
| AI apartment images ban | DCPR consumer protection rule | Realtors and listings |
| Chatbot “doom loop” limits | Drafted city rule | Companion/engagement chatbots |
| Data center construction pause | Supported state legislation | Hyperscale builds |
| Madison Square Garden surveillance | Backing Council member Hanif’s bill | Private surveillance tech |
Job Meaning And The Emotional Cost Of AI
The Professor In Tears
Then Mamdani made an argument rarely heard from an elected official: that the AI jobs debate has a whole neglected chapter, about job satisfaction rather than job counts. His example was The Daily’s interview with a mathematics professor left in tears by AI’s sudden ability to solve an age-old problem that had defined his life’s work — and the crisis of purpose that followed.
There are, in his telling, “a lot of jobs where the very purpose of the job is starting to be stripped away by artificial intelligence.” A spreadsheet that a clerk once built over a week is now a thirty-second prompt; the output survives but the craft dies. Payroll statistics capture none of that — and neither do the siting decisions for new AI data centers, which treat both power and purpose as externalities.
The larger point is a philosophical one aimed at the industry itself: AI is treated as mythical and unknowable, when it is in fact “a thing very much within human control,” built by people with incentives. If it is human-made, then answering what it does to human purpose is not a software roadmap problem — it is a political responsibility, and politics has been mostly absent.
It is also why he ties the emotional ledger to the physical one. The same AI data centers drawing hundreds of megawatts are the infrastructure behind the tools rewriting what professors, clerks and coders can claim as their craft — a community hosts the plant and absorbs the loss of purpose at the same time, while the rewards stay wherever the headquarters are.
Deepfakes, Democracy And The Crisis Of What Is Real
‘Still Weaker,’ And Footage To Prove It
Asked whether technology is making democracy stronger or weaker — the same question WIRED put a year ago — Mamdani answered “still weaker,” and offered evidence from his own inbox. Family members now send him AI videos of himself, believing them real. A relative complimented him on a 45-minute debate with a European member of parliament; he watched it and realised it was not him. (He apparently beat his AI counterpart handily, though he could not sit through the footage.)
The local edition is uglier. Republican gubernatorial candidate Bruce Blakeman — dubbed “Mr. AI” — has run AI-generated ads featuring Mamdani and Governor Kathy Hochul, including one Drummond described as a “delusion pill” skit in which the two discuss how higher taxes somehow feel good. Both Hochul and Mamdani have publicly fought back against the fabricated clips.
Mamdani ties it to institutional trust: in an age of already little faith in traditional institutions, synthetic video can diminish even the memory of what was real. The surveillance thread runs adjacent: WIRED’s reporting on Madison Square Garden’s sprawling surveillance apparatus, and a Council member’s bill to restrain it, drew from Mamdani a flat rejection of private citizens wielding that power over New Yorkers. Technology making democracy weaker is not a dystopia forecast — it is the current inventory.
The connection to his infrastructure argument is direct: AI data centers are the physical base of the synthetic media layer, and the same risk-transfer logic applies. The facilities and the harms are local; the companies that profit from both are not.
What Mamdani Wants Washington To Do
Three Reckonings For Washington
The federal ask comes down to three reckonings: the housing-bubble parallels, the profitability question that keeps the answer vague, and the fact that AI deployments “have already been seen to have fatal consequences” — hallucination harms, chatbot harm, and now fabricated video of candidates on the ballot. The arms-race framing, he argues, has become the universal excuse: “everything that we’re doing is critically important because we are in a race with China” is the sentence used to justify every abdication of duty.
His broader political argument is that local action substitutes for absent federal law — which is both a brag and an indictment. A city can regulate realtor images and school tools; only Washington can price grid externalities, mandate energy disclosures for AI data centers, or set national guardrails on frontier models.
There is a campaign subtext here too, delivered without a shred of subtlety. Mamdani is building a national lane for a party that talks about AI the way it once talked about finance: valuations, externalities, who pays when things break. The midterms will decide whether that language spreads. If it does, the AI data centers conversation migrates from township hearing rooms to Sunday shows — and the companies that spent years perfecting the art of opaque scale will discover what banks learned after 2008: opacity is a loan, and the bill eventually comes due.
The AI build-out economy should measure that political weather as carefully as it measures megawatts. The childcare expansion he celebrates — $1.2 billion lifting Universal 2K from 2,000 to 12,000 two-year-olds next year, with Bronx and Staten Island rollout already scheduled — is what governing with audited numbers looks like. AI data centers will be governed by numbers soon enough; the only open question is whose numbers win.
References
WIRED — ‘Our Risk, Their Reward’: Zohran Mamdani on AI and Data Centers
WIRED — The Big Interview: Zohran Mamdani (October 2025)
WIRED — Madison Square Garden surveillance reporting
The New York Times — AI-generated ads in the midterm elections
The New York Times — Mamdani’s office and creator diplomacy
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
