Big Short Investor Eisman: Local Politics, Not Chips, Is AI's Biggest Risk
Big Short investor Steve Eisman says local and state governments slowing AI data center construction pose a bigger risk to the AI trade than chip supply — and a full moratorium could trigger recession.

Why it matters
- Steve Eisman told CNBC on November 17, 2026, that local permitting, not chip supply, is the biggest AI risk, with a plausible 10–30% construction slowdown.
- The average individual data center load doubled from 150 to 300 megawatts between 2023 and 2024, per American Public Power Association House testimony.
- Amazon spent $77.26B on Q3 capex, up 120%, and expects about $220B for 2026; AWS revenue grew 39%.
- Digital Realty has 1.5 GW under construction with 63% pre-leased; Equinix booked a record $533M in annualized gross bookings.
- Vertiv grew Q3 revenue 61.2% to $4.31B, yet its stock fell 8.85% over the past month and Eaton's fell 13.76%.
"Big Short" investor Steve Eisman told CNBC on November 17, 2026, that the biggest danger to the AI trade is not chip supply — it's local zoning boards and statehouses slowing data center construction. His worst case is stark. A total construction moratorium, he said, "would obviously put the country into a recession overnight." His base case is milder: a 10 to 30 percent slowdown in new builds.
Eisman, who rose to fame betting against subprime mortgages before the 2008 crash, laid out his reasoning in a full CNBC interview. "I think the biggest issues by far are how much are local and state governments going to slow AI data center construction?" he said. "Is it going to be a 100% moratorium? That would obviously put the country into a recession overnight, or is it something like 10 to 30%?"
He also addressed reports about the leading AI chipmaker's sales in Southeast Asia, which he estimated could run around 15% of chip revenue. "That's a lot," Eisman said. Even so, he ranked permitting above chips as the binding constraint. A GPU generates nothing until a powered building exists to hold it.
That framing matters for the AI industry because the entire buildout — hyperscaler capex, power equipment orders, REIT leasing pipelines — depends on a permitting regime controlled by thousands of local governments.
Why Are Towns Pushing Back on Data Centers?
A moratorium is a temporary pause on new permits, usually while a town or state rewrites its rules on zoning, power, or water. The resistance comes down to three things: electricity, water, and land.
A Colorado legislative staff memo notes that data centers require large amounts of electricity for computational power and water to cool the computer systems. It adds that laws covering land use, water use, and protections for residential customers have already passed nationwide.
The load numbers explain the friction. In House testimony, the American Public Power Association said the average individual data center load doubled from 150 to 300 megawatts between 2023 and 2024. The EIA says U.S. electricity demand has grown 2.1% per year on average over the last five years, after 15 flat years.
The pressure is already redirecting projects. On Digital Realty's (NYSE:DLR) third-quarter call, Chief Investment Officer Gregory Wright pointed to "numerous moratoriums across Michigan" and power queues in Ohio as reasons hyperscalers are moving into Indiana.
Who Takes the First Hit From a Slowdown?
Data center REITs. Digital Realty has 1.5 gigawatts under construction, with 63% already pre-leased. Its chief executive acknowledged: "Yes, there's always here and there a delay." At Equinix (NASDAQ:EQIX), CEO Adaire Fox-Martin said: "Power availability, legislation and community sentiment have created some headwinds for the industry." She also noted that cancellations would make "already scarce capacity much more of a scarce resource." Equinix stock is up 36.94% year to date.
Power equipment. Vertiv (NYSE:VRT) grew Q3 revenue 61.2% to $4.31 billion. Its CEO said permitting and power are "pacing items, but they are not slowing down growth." Eaton reported its Electrical backlog up 46%. Investors still look nervous. Over the past month, Vertiv has fallen 8.85% and Eaton 13.76%.
Power producers. Constellation Energy (NASDAQ:CEG) is down 22.76% year to date. Its CEO called that "a bit of an overreaction in the market to the election results based upon fear that Democrats won't let data centers get built."
Why Could Amazon Win From Slower Construction?
Eisman's second argument flips the risk on its head. "But longer term, I think if the AI data center construction story is 10 to 30% less, the group that makes out the best are the hyperscalers because they go cash flow positive," he said. He added that a slowdown would make construction "much more sustainable."
Amazon (NASDAQ:AMZN) is the clearest case. The company spent $77.26 billion on capex in Q3, up 120%, and still expects about $220 billion for 2026. AWS revenue grew 39% over the same period. If building slows, that spending falls while tight capacity supports pricing — a combination that flatters the cash flow statement.
The market logic is straightforward. Less construction means fewer dollars out the door for hyperscalers, and scarcity means existing capacity prices better. Suppliers that depend on construction volume, like Vertiv and Eaton, face the opposite arithmetic.
Is Eisman Watching the Right Bottleneck?
The earnings data back him up on demand. Equinix booked a record $533 million in annualized gross bookings. Vertiv's organic orders more than doubled. The weak link in the chain is getting buildings permitted and energized — not finding customers for them.
That bifurcation defines who wins under each end of Eisman's range. Suppliers like Vertiv and Eaton depend on construction volume. Landlords that already hold permitted, energized sites, and hyperscalers paying the capex bill, could come out ahead if capacity tightens.
Three markers will show which end of the range plays out:
- FERC's ruling on colocation. Constellation expects it in the first half of next year.
- Equinix's Q4 deliveries. The question is whether it delivers more than 13,000 cabinets.
- Digital Realty's pre-leased capacity. The REIT has 250 megawatts of pre-leased deliveries due over the next two quarters.
Each signal tests the same thing: whether demand keeps converting into energized buildings, or whether politics stands between the GPU orders and the grid.
The stakes run past investor portfolios. Eisman's recession scenario — a 100% national moratorium — remains unlikely, but the 10 to 30% band he calls plausible would reshape capital allocation across the AI supply chain. Watch Michigan, Ohio, and FERC's calendar. That's where the AI trade's real bottleneck lives.
Source: 247wallst.com
More from Elena Vasquez
Show full bio
Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.
708 articles
Related articles
- Michael Burry Says Nvidia's Real Problem Is a $176 Billion Accounting Gap
- Anthropic Targets $1 Trillion to $1.5 Trillion IPO as Voters Reject Data Centers
- Midterms widen the AI divide: CoreWeave falls as hyperscalers gain
- Sam Altman Warns Chip Suppliers Are All Counting the Same Buyers
- David Sacks Warns the AI Data Center Pipeline Could Close by 2028