Data Center Moratoriums: How Worried Should Investors Actually Be?

Key Takeaways

• More than 300 data center restrictions do not translate into a comparable amount of delayed capacity. The research attributes about 2.3 GW of delay to local and state actions.
• Moratoriums appear to be adding friction and favoring projects with secured power, permits and community support, rather than stopping the U.S. buildout.
• Investors can track whether restrictions reach already-entitled projects and whether planned capacity converts to construction to gauge how the risk is changing.

Over the past year, a specific kind of headline has become hard to avoid:

• Another town has passed a moratorium on new data centers.
• Another state has issued an executive order pausing approvals.

Commentators have started adding up the count, more than 300 such restrictions, by one widely cited tally, and using that number to argue that the artificial intelligence infrastructure buildout is running into a wall.

For investors, this is not an abstract concern. Spending commitments for AI infrastructure, including land, power, chips and networking equipment, have generated enormous order backlogs across a range of industries. That backlog, more than any single quarter’s revenue, has been moving equity markets. If the buildout is genuinely at risk of stalling, that matters to portfolios positioned around it. If the risk is overstated, or is real but far narrower than the headlines imply, that also matters, but in a different way.

The useful question is not “are moratoriums happening?” They clearly are. The useful question is what fraction of the planned capacity they actually touch, and how a partial slowdown differs from a stoppage.

Why Counting Moratoriums Is the Wrong Metric

A research note published in September 2026 by SemiAnalysis¹ attempted to answer this question directly by mapping several hundred local and state restrictions against a database of thousands of data center projects, using permitting records, property records, power interconnection data and satellite imagery. Rather than treating each restriction as equivalent, the analysis tested whether a given restriction actually overlapped with a specific project in a way that could plausibly delay it.

That overlap turns out to be the exception rather than the rule. A restriction only matters to a project if:

• reaches the exact parcel
• is active when an approval is needed
• covers the proposed use
• blocks a step that is still required
• has no applicable carve-out
• sits on the project’s critical path

Missing any one of those conditions is often enough to reduce the modeled delay to zero. The research finds that roughly 80% of enacted local restrictions contain no modeled planned capacity at all because many were adopted preemptively or after the relevant approvals had already been secured.

The scale of the gap between exposure and delay is the piece’s most important number.

• Of roughly 20.1 gigawatts (GW) of capacity within the boundaries of live local restrictions, the analysis attributes a directly modeled delay to only about 1.5 GW, or roughly 7.6%.
• Adding the effect of New York’s statewide permitting pause brings the directly attributable total to approximately 2.3 GW.

Compared with a 2027 U.S. net-addition forecast of 38 GW, of which 22 GW is already under vertical construction, the 2.3 GW impact is modest and geographically concentrated. It is not evidence of a national slowdown.

The Spectrum Between “Harder” and “Stopped”

None of this means the political backlash is irrelevant, and treating it that way would be its own analytical mistake. There is a meaningful difference between a buildout becoming more difficult to execute and a buildout actually stopping, and the evidence, at least so far, supports the former, not the latter.

State-level actions illustrate how much this varies by jurisdiction, and why lumping them together under “moratorium” obscures more than it reveals.

• New York’s executive order pauses discretionary environmental permits for new facilities above 50 megawatts. It affects roughly 1.4 GW but is estimated to delay closer to 0.8 GW, shifting a handful of projects by two to three quarters.
• Texas’s grid operator has paused new large-load connection approvals while it verifies eligibility. The research characterizes this as adding a few months of administrative delay for grid-connected projects, while leaving behind-the-meter projects, which bypass the grid queue, largely unaffected.
• Pennsylvania’s order does not impose a temporary freeze. It removes data centers from fast-track permitting and lengthens diligence, raising the bar for speculative projects without necessarily changing outcomes for well-capitalized ones.
• Oregon’s pause applies only to transactions involving state-owned land. It occurs upstream of the point at which a project would register in a capacity forecast.

The four states have different mechanisms and different degrees of impact. That is difficult to compress into a single “moratorium wave” narrative, and doing so is what produces overstated fear.

There is also a useful causal read hiding in the data.

Moratorium activity looks less like a leading indicator of collapse and more like a lagging indicator of visible growth, and, in turn, a leading indicator of rising friction.

In Ohio, data center capacity nearly tripled, from about 1 GW in 2023 to 3 GW in 2025, before roughly 40 local moratoriums followed. The backlash came after the buildout became visible to nearby communities, a very different dynamic from backlash preventing a buildout from happening in the first place.

What the Evidence Does Not Prove

It would be just as much of a mistake to swing too far toward complacency. The research is explicit that its findings describe the present, not a guarantee about the future. Moratoriums could become more material if they grow more widespread, more durable, or more restrictive than they are today. Public sentiment is not improving on its own. An August 2026 survey cited in the research found 46% of respondents viewed data centers unfavorably, and 46% opposed a new facility in their own community. And the reassuring 38 GW 2027 forecast is not risk-free in its own right. 16 GW of that figure remains classified as “planned” rather than under vertical construction, even though much of it is financed and already in site work.

The key distinction is between delays caused specifically by local policy and the broader set of constraints that can independently slow a project, regardless of what any town council decides. These include grid interconnection queues, equipment lead times, financing, litigation and ordinary construction timelines. Moratoriums are one input into project timing, but they are not the only one or currently the dominant one.

What This Means for Positioning

If the evidence supports friction rather than stoppage, the more useful investment framework may be to treat the moratorium wave as a story about selectivity rather than aggregate demand. Restrictions that filter out speculative or undercapitalized projects tend to raise the relative value of what remains. Sites with existing entitlements and industrial zoning, developers with secured power and community agreements already in place, and operators with the balance-sheet capacity to absorb longer diligence timelines and community-benefit negotiations. Grid-level restrictions, in particular, appear to be accelerating interest in behind-the-meter and bridge power, including gas turbines, reciprocating engines, fuel cells, and on-site generation more broadly, as a way of sidestepping interconnection queues rather than a sign that total computing demand is shrinking.

In other words, aggregate demand for AI infrastructure can persist even as returns diverge sharply across individual projects, developers, and suppliers, depending on how well each one navigates the permitting and power landscape that is now visibly harder than it was a few years ago.

The Honest Bottom Line

The popular narrative that more than 300 moratoriums are shutting down the U.S. data center buildout does not hold up against project-level data. The more accurate narrative is narrower and more useful.

A small, quantifiable share of planned capacity is directly delayed by local restrictions. A much larger share of the political reaction adds friction that reshapes where and how projects get built, rather than whether the buildout continues.

That does not mean investors should dismiss the political risk. It means the risk should be sized correctly and tracked through specific indicators: whether restrictions begin reaching already-entitled projects, whether temporary local actions become durable statewide laws, and whether the “planned” portion of near-term forecasts converts into construction. Headline counts of ordinances passed are less informative. The buildout is not stopping. It is becoming more selective about where it happens and who is positioned to execute it. That distinction should shape how investors approach the opportunity.

What a Welcomed Data Center Looks Like

Not every community resists data centers, and the contrast is worth considering. In a September 2026 episode of the All-In podcast, Meta president Dina Powell McCormick discussed the company’s Hyperion campus in Richland Parish, Louisiana, with Louisiana Economic Development Secretary Susan Bourgeois and Richland Parish Schools Superintendent Sheldon Jones. Meta committed $10 billion to the site in December 2024. By July 2026, that commitment had grown to more than $50 billion for a campus approaching 10 million square feet and 5 GW of capacity.

The economic effect has been concrete. As construction activity ramped up, local sales tax collections, which typically grow 5% to 10% a year, rose 60% to 70% and briefly peaked near 260%. Because a portion of that tax funds teacher pay, the effect showed up directly in paychecks.

A certified teacher’s maximum supplemental check rose from roughly $10,200 to more than $50,000 in a single year.

Jones described how a parish that once struggled to recruit teachers now fields calls from educators across the state, offering young residents a chance to stay home for work.

Richland Parish does not contradict the friction documented above; it illustrates the other side of the same picture. Projects with secured power, credible local agreements and visible, broadly shared economic benefits are precisely the ones the moratorium wave is not stopping. Increasingly, they are the projects communities are competing to attract.

Originally posted on October 7, 2026 on WisdomTree blog

FOOTNOTES AND SOURCES

¹ Barkin, M., Knuhtsen, R., Eliahou Ontiveros, J., & Patel, D. (2026, September 15). Everyone says datacenter moratoriums are killing the US buildout. We disagree. SemiAnalysis.

DISCLOSURES