skip to main content

Data Centers Are Spooking Americans

Data Centers Are Spooking Americans

Electricity use is the biggest concern

by Ransi Clark, Alex Halsey, and R. Michael Alvarez, California Institute of Technology

May 18, 2026



Recently, it seemed that the first-ever state ban on hyperscale data centers might be enacted. Maine's 18-month moratorium had moved through both chambers, but was scuttled by the governor's veto. Governor Mills said she would have signed it if it included an exemption for the data center in the former Androscoggin Mill site in the town of Jay, which the town had already approved. After the mill closed in 2023 from an accident, the town had trouble repurposing it, but a data center provider had been signed on.

Elsewhere in the United States, data centers are replacing a variety of dead infrastructure. After the retail chain Sears filed for bankruptcy, their headquarters in Illinois was bought up by Compass and will now become a 200 MW data center campus. In Kentucky and Maryland, idled aluminium smelters are being transformed into data centers.

Despite the promise of this second industrialization, the broader American public treat data centers with great skepticism. Over 50 counties have banned them including 12 in Michigan, 11 in North Carolina, and 9 each in Ohio and Georgia. Almost all these bans are short term, but some places like Vermont's Royalton have 5-year bans in place, and California's Monterrey Park is considering a permanent ban. Even when local governments don't outright ban them, several projects are not approved and cancelled. We counted at least 88 such cancelled projects.

To understand the nature of this opposition to data centers, we conducted a survey through YouGov in March 2026. The survey asked 2,187 respondents their awareness of AI chatbots and their preferences for data centers. More specifically, the respondents were asked about their support for data center moratoria at various levels of government followed by what their major concern about data centers were. Figure 1 summarizes the results. All presented estimates are weighted to match the YouGov national profile.

Figure 1: Extent of support for data center bans at national, state, and local levels.

Respondents were first asked how much they support a national ban on data centers. The most common response (45%) was "not sure," but 35% of the sample supported a national ban, while 20% opposed. Those who said they were "not sure" or opposed a national ban then were asked about their support for state-level bans: 200 of those not sure on the national ban then supported a state ban, and 345 of those who opposed a national ban also opposed a state ban. This means the support for either a state or national ban on data centers to 46% of the sample (second blue bar in Figure 1).

County bans were more popular than the state bans, but not by much. About 90 not sure respondents on the state ban, then supported a county ban, bringing the total support for any type of ban to about 52% (third blue bar in Figure 1).

The respondents were then asked what their support for a county data center was. The opposition for a local data center was below 50%, and support for a local data center was 32%, about twice as high as the opposition to a county ban. About 10% of the sample (225 respondents) supported one of national/state/county bans, but also supported a local county data center. These respondents may have misread one of the questions, or did not grasp that a national/state/county ban would mean a county data center will not be built. Even with such a confusion, opposition is the most common reaction.

The respondents were then asked what their support for a local data center was. The opposition was comparable to those who supported a county ban, but the support for a local data center is 31%, somewhat higher than the opposition to a county ban. About 225 respondents that supported a county ban also ironically supported a local data center. The contradictory choices might be due to how respondents interpreted what local meant, given that counties are of varied sizes across the United States.

There is considerable geographic variation in the preference for a state ban. Figure 2 illustrates this variation. In all states except Wyoming and Alabama, the support for a state ban, outnumbered opposition to ban. This includes states such as Ohio, Texas, and Nevada that are seeing rapid data center buildouts.

Figure 2: Support for a state ban at a state level.

Even with similar levels of opposition, there is great variability in political action. For example, Texas and Michigan both have similar levels of support for a state ban, but only one county in Texas has a ban as compared to 12 counties in Michigan. State level data center bans are moving through at least 18 state legislatures, but many are stalled in committee and will never be voted on. State officials are likely aware that data center bans are popular among the public, but are feeling that this should be a local issue. Instead of outright bans, state governments are rolling back some of the tax incentives and looking for assurances that the utilities and consumers won't be holding the bag.

Across all four questions, it is clear that a sizable portion of the American public is wary of data centers. What is driving this wariness? Early reporting on data center opposition implied that local opposition is driven by worries about its freshwater use. Several high-profile data center projects such Amazon's Project Blue and Google's Project Flo were cancelled or delayed specifically because of water use.

When asked what their biggest environmental concern was, however, more than 45% of the sample identified their major environmental concern to be electricity use by data centers (See Figure 3). About 28% identified water use as the major concern, and 5% said their major concern was noise. Perhaps those initial concerns about water were superseded by concerns about data center electricity demand.

Figure 3: Major concerns about data centers.

Though concerns about water are possibly exaggerated by misleading early reporting, the public is right to worry about data center electricity use. Data center operators concede that getting sufficient supply is their major bottleneck. Some operators are resorting to onsite energy generation, but gas turbines are also in short supply. Hyperscalers are now buying entire nuclear plants just to meet the AI computational demands of their data centers.

Hyperscalers are not done yet, either. More and more Americans are becoming regular users, and demand for computation will increase commensurately. Vacancies in data centers are at an all-time low while rental rates for data centers are rising. Many hyperscaler providers are diversifying to new markets with the largest ever data center approved in Utah just this week. Others are hoping to avoid data center politics entirely and shoot them into space.


Interested in subscribing to receive LCSSP's latest updates?

If you are external to Caltech, fill out this form.

If you are on the Caltech network, use these links to subscribe: LCSSP Mailing List, LCSSP BioPolicy Initiative, and LCSSP Democracy Mailing List.

For help or questions, please reach out to [email protected]!