Unscarcity
Sign in for free: Preamble (PDF, ebook & audiobook) + Forum access + Direct purchases Sign In By signing in, you get our daily podcast email. Unsubscribe any time.

Unscarcity Research

Data Center Moratoriums: Who Vetoes the AI Buildout?

New York froze big data centers; Texas paused a 474 GW queue; 116 towns moved first. When democracy can veto the AI buildout, who decides what gets built?

17 min read 3886 words Updated September 2026 /a/infrastructure-veto-points

Note: This is a research note supplementing the book Unscarcity, now available for purchase. These notes expand on concepts from the main text. Start here or get the book.

Data Center Moratoriums: Who Vetoes the AI Buildout?

The AI buildout has a physics problem and a politics problem. The politics problem is winning.

On July 14, 2026, Governor Kathy Hochul signed Executive Order No. 62, making New York the first state in the country to slam a moratorium on large data centers. Any project drawing 50 megawatts or more, the appetite of a small city, now has its state environmental permits held in abeyance until regulators finish a Generic Environmental Impact Statement, a job the governor described as taking about a year. The order itself sets no deadline.

The context makes the move less surprising than it sounds. As of May 2026, roughly 12 gigawatts of data center load, more than eight of it having piled into the queue in 2025 alone, was waiting to plug into New York’s grid. That is not a rounding error. That is a second New York City’s worth of demand, showing up all at once, to run machines that employ almost no one locally and export their value to shareholders somewhere else. Chips got cheap and capital got abundant, but the right to say no turned out to be its own scarce layer, and a zoning board just found out it’s sitting on top of it.

New York got the headline. But it was late. By the end of June 2026, 116 municipalities had already imposed their own local moratoriums. Denver’s city council voted unanimously in May to pause new data center permits for a year. Seattle rushed through an emergency moratorium after constituents sent 93,270 emails opposing five proposed campuses. Lawmakers in fifteen states filed bills that would impose one in the same legislative season. The most abundant infrastructure of the intelligence age is running headlong into the oldest machinery in a democracy: the right to say no.

This is the third fight in a trilogy. We’ve covered who pays for AI’s electricity, the ratepayer battle over cost. We’ve covered who owns the reactors that generate the power, the sovereignty battle over the machine. This is the one underneath both: who gets to decide whether the thing gets built at all, and where.

The Veto Went Mainstream

For most of the last decade, data centers were invisible. They arrived in tax-abated business parks, made a low hum, and nobody outside the county assessor’s office cared. That truce is over, and the polling tells you why.

A 2026 Gallup survey found that 71% of Americans would oppose a data center being built near them, 48% of them strongly. To put that number in perspective: opposition to a data center now runs higher than opposition to a nearby nuclear plant, which clocks in at 53%. People are more frightened of a warehouse of GPUs than of a fission reactor. That is a remarkable thing for the abundance story to have to reckon with, and it did not happen by accident.

The objections are concrete, and mostly reasonable. A hyperscale campus can draw millions of gallons of water a day for cooling. It pushes up the wholesale price of electricity for everyone on the same grid. It runs cooling fans that turn a quiet exurb into something that sounds like a permanent freeway. It occupies land, alters the water table, and, in exchange, hires a security guard and a few dozen technicians. When residents of Tonawanda, New York packed a public meeting to fight a proposed campus, they weren’t reciting talking points. They were doing the arithmetic on their own well water and their own property values and finding that the deal, as offered, was lousy.

The result is a permitting environment that has become genuinely treacherous for the buildout. One industry tally counted at least 75 projects worth a cumulative $130 billion delayed or blocked by local opposition in the first quarter of 2026 alone. The bottleneck on AI, it turns out, may not be chips or even electrons. It may be a zoning board in a town you’ve never heard of.

By late summer the veto had become electoral arithmetic. On August 31 the president posted that a community refusing a data center wants to end up “backwards and poor” and signed off with “let Data Reign.” Hours later his own vice president read the bill instead: “probably 99% of the backlash to data centers has come in areas where building a data center means higher utilities and higher electricity for the people on the ground.” A National Republican Senatorial Committee memo dated August 18 warned that the campaign against data centers “will expand far beyond Ohio.” AdImpact counted more than $31 million in political ads mentioning data centers this year, more than 99% of it against, from both parties. And in Ohio, which has the fourth-largest fleet of data centers in the country, voters in 18 municipalities and townships will decide bans and zoning caps on the November 3 ballot, nine of them modeled on a statewide 25-megawatt ban that missed its signature deadline this year and is aiming for 2027. The veto has moved from the zoning board’s agenda to the ballot itself, by name. The J-curve article walks through what that costs politically; this one is about who holds the lever.

What a Veto Point Actually Is

“Veto point” is political-science jargon for a boringly powerful idea: any stage in a process where one actor can stop the whole thing. A bill dies in committee. A permit sits unsigned. A single county commissioner tables a rezoning. The AI buildout has to pass through a startling number of these, and each one is a place where “no” costs almost nothing and “yes” costs a career.

Stack them up and you get the map of who can slow the machine:

  • Local zoning and land use. The town decides what can be built where. This is the oldest and most decentralized veto in American government, and it is where 116 of those moratoriums live.
  • Discretionary state permits. Environmental review, water withdrawal, air quality. New York’s moratorium works by freezing exactly this layer, the state permits a project needs before it can break ground.
  • Grid interconnection. The utility and the grid operator decide when, and whether, you get to plug in. A multi-year queue is a veto that never has to speak its name.
  • Environmental impact review. The requirement to study, disclose, and mitigate. New York’s pause explicitly runs until the state finishes a Generic Environmental Impact Statement, turning the study itself into the brake.

In the two months after New York moved, three of those four layers fired at the state level, each in a different register. Texas, the most permissive grid in the country, pulled the interconnection lever. On August 3 Governor Greg Abbott ordered an audit of every data center in ERCOT’s queue, a queue that had swollen to 474 gigawatts of requests, more than five times the grid’s all-time peak, and warned that projects failing the check “will be denied connection.” ERCOT is surveying 461 data centers with a 130-question form and will not finish before December 10; one industry estimate puts up to a fifth of the entire US data-center pipeline on hold behind it. Pennsylvania pulled the permit lever. Governor Josh Shapiro’s Executive Order 2026-05 of August 18 makes state environmental review of any data center above 25 megawatts wait for local land-use approval, throws every data center out of the state’s permit fast-track, and bans the non-disclosure agreements developers use to keep a town in the dark until the deal is signed. Maine showed that the veto has a veto of its own: its legislature passed the first statewide moratorium in the country, running to November 2027, and the governor vetoed it because it failed to exempt one project with strong local support. New York’s own legislature had passed a stricter 20-megawatt version back in June; it sits on the governor’s desk until December 31 and dies by pocket veto if nothing is signed.

Here is the uncomfortable part for anyone who loves the abundance vision: these are the same veto points that block housing, transmission lines, and clean energy. The environmental review that stalls a data center is the environmental review that stalls the solar farm meant to power it. The zoning board that says no to GPUs is the zoning board that said no to the apartment building. You cannot cheer the veto when it blocks the thing you dislike and boo it when it blocks the thing you need. It’s one machine, and it doesn’t know the difference.

That is the tension the Unscarcity framework has to actually resolve, not wish away. A civilization that guarantees energy and compute as part of the Foundation has to build the generators and the data halls that make the guarantee real. A civilization that lets any locality veto any project builds nothing. But a civilization that overrides local objection by fiat has just reinvented the thing the framework exists to prevent: infrastructure imposed on people who never consented, for the benefit of owners they’ll never meet.

The Legitimate Core, and the Capture

So untangle it. Some of that 71% is pure NIMBYism, the reflexive “not in my backyard” that would block a bakery if it meant a new parking situation. But most of it is something sharper and more defensible: a rational response to a genuinely bad bargain.

Run the deal a hyperscaler offers a typical town. The company captures the value: the compute, the model, the revenue, all of it flowing to a balance sheet in another state. The town eats the cost: the water, the noise, the higher power bills, the strain on a grid it shares. The jobs are few and the tax abatements are generous. Stated plainly, the community is being asked to subsidize a trillion-dollar company’s private infrastructure and receive a hum in return. The moratorium isn’t irrational. It’s a negotiating position, and a sound one.

Which means the veto is doing something useful. It is the mechanism that forces the conversation the cost article already framed: make the beneficiary pay, and make it share. A project that internalizes its own costs (brings its own generation, funds the water infrastructure, pays a real host-community benefit, prices its grid impact into its own bill rather than yours) is a project most of that 71% would actually wave through. The objection isn’t to compute. It’s to being fleeced. Remove the fleecing and you remove most of the veto’s fuel.

This is precisely the logic of the land tax that funds abundance: capture value where it accrues, return it to the community that hosts it. A data center that made the town richer instead of poorer wouldn’t need to override the town’s objection. It would be welcome.

The Veto Gets Tested

Two months is long enough for the first counter-moves. In California’s Imperial Valley, where the county board froze all data-center development until mid-2027 after months of packed hearings, the developer sued, and in August a superior court judge issued a tentative ruling that the county had failed to identify a “current and immediate threat” to public health, safety or welfare, the legal threshold for an emergency moratorium. “They identify concerns and the possibility of conflict with regulations still being studied,” the judge wrote, “but not the current and immediate threat required.” The veto, it turns out, has procedure of its own. A pause has to be justified on the record, not merely voted, and the developers have lawyers too. Expect the arbitrage described in the next section to run through courtrooms as much as across county lines.

The more telling test is in the place with the least excuse to say no. Loudoun County, Virginia, is the densest data-center market on earth, and by the industry’s own accounting it cut property taxes and expanded public services on data-center money for a decade. In September its Board of Supervisors scheduled a vote on pausing new applications while it finishes an electrical-infrastructure plan and stricter noise rules. “We’ve become addicted to the data centers for their tax revenues,” the supervisor who proposed the pause told reporters, “but at what cost?” When the county that got rich on the buildout starts asking that question, the objection is no longer a bad deal for a poor town. It is saturation, and saturation is the one problem a better host-community agreement cannot fix. Coastal Georgia is running the same script: Chatham County, home to Savannah, enacted a 155-day moratorium on anything over 10 megawatts on September 11, the third county on that coast to do so, days after word spread that a $20 billion OpenAI campus in the next county over had been negotiated in secret.

The Failure Mode: The Buildout Just Moves

But watch what happens when the beneficiary refuses to share, and the veto stands anyway. The machine doesn’t stop. It relocates.

Block a data center in Westchester and it reappears in a rural parish in Louisiana that can’t afford to say no, or a county in Wyoming that’s desperate for the tax base, or across a border entirely. Capital is mobile; a server rack has no homeland. This is regulatory arbitrage, and it is the ugliest possible outcome, because it means the veto held only for the communities with the power to hold it. The wealthy suburb gets its quiet. The poor county gets the cooling towers, the truck traffic, and the strained aquifer. The buildout still happens. It just gets sited on the people with the least ability to resist.

We’ve seen this movie. It’s the story of every landfill, every refinery, every incinerator in the history of environmental justice: the externality flows downhill to wherever resistance is cheapest. A veto point that only functions for the strong isn’t democratic control of infrastructure. It’s a sorting mechanism for dumping the costs on the weak.

And it scales up to the geopolitical level. If one nation’s environmental review makes the buildout slow and expensive, the compute migrates to a jurisdiction that treats a permit as a formality. The veto, exercised locally, quietly hands the AI frontier to whoever is willing to bulldoze their own citizens’ objections fastest. Democratic friction in one place becomes a strategic gift somewhere else. This is the same trap the book’s Sovereign EXIT analysis keeps circling: a race where the least-restrained player sets the pace for everyone.

What the Book Actually Proposes

The framework’s answer is not “override the vetoes” and it is not “let every town veto everything.” It is to change the deal so the veto rarely has to fire, and to build a governance layer that can tell a legitimate objection from a captured one.

Three moves, all of them already in the book:

Make the objection dissolve by internalizing the cost. The MOSAIC, the framework’s federated governance system, doesn’t grant infrastructure the right to steamroll a community. It grants communities the standing to demand that projects carry their own weight, in water, power, noise and benefit, before they’re built. A veto point, in this reading, is not a bug in the buildout but the pressure that converts an extractive project into a shared one. Keep the veto; remove the reason to use it.

Guard against capture in both directions. The Diversity Guard exists because no single faction, neither a captured regulator waving projects through nor an organized NIMBY bloc freezing all development, should be able to seize the process. Real governance has to distinguish the town protecting its aquifer from the town protecting its property values from an apartment building. That’s a judgment the MOSAIC is explicitly designed to make: legitimacy requires assent across genuinely different communities, which is hard to fake in either direction.

Build where you’re wanted, on ground where nobody has to lose. This is the deepest move, and it’s why Free Zones start on greenfield. The whole reason the land-tax funding model works on empty ground is that there are no entrenched interests to override: no existing homeowners, no prior water claims, no veto to bulldoze. A community built from scratch, by settlers who chose to be there for the surplus it offers, doesn’t have a siting problem. It is the site. The Free Zone sidesteps the veto war not by defeating the vetoes but by building somewhere they were never cast.

Notice that none of these is “trust the buildout.” The framework is not on the side of the hyperscalers here. Infrastructure libertarianism, the book’s argument that public infrastructure expands freedom rather than shrinking it, cuts hard against the company-town model, where a private owner captures a place and its politics. A data center that turns a county into a subsidized colony of Azure is not abundance; it’s the Star Wars timeline the Preamble warns about, arriving one permit at a time.

The Honest Bottom Line

New York’s moratorium will be portrayed two ways, and both will be wrong. The buildout crowd will call it Luddism, democracy getting in the way of progress. The degrowth crowd will call it a victory, the little guy stopping Big Tech. Neither reads the situation honestly.

What actually happened is that a governance system caught up to a physical reality faster than expected, and discovered it had a lever. The lever is real, and it is legitimate, and it is also dangerous, because a veto that only works for the strong is worse than no veto at all, and a buildout that routes around every objection ends up sited on whoever is too poor to object.

The question the book keeps asking, in its energy chapters and its governance chapters alike, is whether abundance gets governed or merely absorbed. The veto points are where that gets decided, in real time, in county meeting rooms and state permit offices most people will never see. Handled well, they’re the mechanism that forces the AI buildout to pay its way and share its surplus. Handled badly, they’re just a filter that sends the costs downhill and the compute overseas.

The moratoriums are not the end of the buildout. They’re the first time the buildout has been asked a question it can’t answer with a check. What’s in it for the people who have to live next to it? That’s not an obstacle to abundance. That’s the entire point of it.

Further Reading

  • The Abundance J-Curve — Why the buildout raises bills before it lowers them, and the electoral arithmetic that turned the veto into a campaign issue.
  • Who Pays for AI’s Electricity? — The cost fight: the sibling battle over whose power bill funds the buildout, and the “make the beneficiary pay” principle this piece extends to siting.
  • Own the Reactor, Own the Compute — The ownership fight: who holds equity in the machines that generate the power the data centers need.
  • The Electron Gap — Why energy, not chips, is the real bottleneck — and why a siting veto bites so hard.
  • The MOSAIC Architecture — The federated governance system that grants communities standing without granting anyone a monopoly on “no.”
  • The Diversity Guard — How to tell a legitimate objection from a captured process, in either direction.
  • Free Zones — Building on greenfield to sidestep the veto war entirely.
  • Settler Economics — Why communities built by people who chose to be there don’t have a siting problem.
  • Land Tax Funds Abundance — Capturing value where it accrues: the funding logic that dissolves most siting objections.
  • The Geopolitics of Abundance — What happens when the buildout routes around any jurisdiction that says no.
  • Infrastructure Libertarianism — Why public infrastructure expands freedom, and why the company-town model betrays it.

References

The AI buildout can buy the chips, buy the power, and buy the reactor. The one thing it can’t buy outright is a community’s consent. That’s not a bug in abundance — it’s the price of admission. Argue about it at unscarcity.ai/forum.

Share this article:

Share Share Share