Author: Brian Merchant; Compiler: BitpushNews

The data center protest movement has swept across the United States. Communities from Vermont to Oklahoma, Indiana to California are organizing to stop the tech industry from ramping up data centers on their doorsteps. This week, the New York State Legislature passed a one-year moratorium on data center construction and is on its way to the governor’s desk for signature. The governor of Chicago (Illinois) has also suspended tax incentives for data centers. Few issues can generate as much political excitement or cohesion as this one; the rare consensus displayed by both parties in the United States in 2026 is the disdain for data centers and hostility to artificial intelligence (AI).

This article will bring an exclusive report from a data scientist. Through in-depth data mining, he accurately analyzed who is blocking the construction of data centers and how successful these protests are.

If you think I'm exaggerating, take a look at a survey just released by Heatmap. The survey asked more than 4,000 Americans about their attitudes toward data centers and whether they support such projects near their homes.

The results show that the public’s negative sentiment towards data centers has become irreversible. Polls show that55% of Americans are “strongly” opposed to building data centers in their area. This is a "record low, revealing a surprising shift in public opinion toward the facilities that underpin the AI boom."

Group portraits of opposing voices

Opposition is particularly strong among Democrats, residents of rural areas, and young people: among those aged 18 to 35, a whopping 80 percent oppose data centers. (This is entirely consistent with the general trend of public opinion; other polls and countless anecdotes have long confirmed that Gen Z has a deep-seated hostility to AI. Just look at the boos that erupted in the graduation speeches that promoted AI this summer.)

But, as readers are well aware, there has been debate and questioning about the driving force and nature of this growing boycott.

Some have argued categorically that opposition to data centers is nothing more than conservative NIMBYism (“don’t build in my backyard”) and is led by environmentalists wearing Patagonia clothing and living in comfortable conditions. While the sheer number of Americans opposed to data centers shown in the Heatmap survey suggests otherwise, there is no data from the poll specifically examining these class factors.

If you want to counter that argument—as Astra Taylor, I, and Saul Levin do—by arguing that the data center opposition movement is actually rooted in working-class politics—then having solid data to back it up is even more important, and that’s where data scientists come into play. After I published a story on the “data center rebellion,” which relied on my first-hand interview transcripts and combing through national news, researcher Geoff Holtzman contacted me to share his quantitative analysis of the movement, which centered on looking at who was actually protesting.

Holtzman describes himself as "a philosopher and data scientist who writes about quantitative propaganda and the rhetoric of scientism," and he frequently publishes articles in his Science & Power newsletter. His peer-reviewed work has been published in prestigious journals such as Proceedings of the National Academy of Sciences (PNAS) and the American Journal of Bioethics. He also heard the widespread theory that the data center protest movement was led by wealthy NIMBY groups, and decided to find out. He conducted a comparative analysis of current and proposed data center project data sets with U.S. Census data (Note 1) and agreed to share his findings exclusively here. He draws at least three striking conclusions.

1. Resistance to data centers in the poorest neighborhoods is nearly five times higher than in the highest asset neighborhoods (19.0% vs. 3.8%).

(Chart description: These quartiles are calculated only for census tracts in the data center dataset, not for the entire United States.)

“The most frequent resistance comes from communities where the median household income is between $8,000 and $72,000,” Holtzman noted. "The communities with the lowest resistance rates have average annual household incomes between US$133,000 and US$250,000."

This directly debunks the political myth that the data center opposition is led by the affluent, Patagonia-wearing upper-middle class; poor or blue-collar communities protest far more frequently than the wealthy.

As Holzman said: "Putting aside any questions of morality or justice, just from a prudential perspective, it would actually be much easier for technology companies to build computing centers in higher-income areas."

He added: "Among low-income, low-educated communities facing project proposals, resistance has been most intense among communities with the lowest incomes and least education." Meanwhile:

Highly educated, high-income neighborhoods showed unusually little protest. In terms of the role that homeownership might play, we're not talking about old money resisting affordable housing—we're talking about precisely the people who might be living in affordable housing.

Moreover, Holtzman’s data confirms that the data center revolt is working. We've seen plenty of headlines about development projects being canceled or scaled back - just this week, Ken O'Leary's mega-project in Utah was cut in half by the state's governor, amid intense public pressure. Other projects were canceled outright.

According to Holtzman’s analysis:

2. Proposed data centers that have faced recent boycotts are more than five times more likely to be canceled or suspended than data centers that have not faced protests (28.2% vs. 5.2%).

This is a very astonishing number. When newly proposed data center projects encounter community resistance, nearly one-third of the projects are ultimately canceled, suspended, or closed. This is an extremely significant success rate, and it should provide further inspiration to those organizers of the data center opposition movement who are weighing whether to launch a new round of resistance.

Finally, combining the insights from the previous two points, Holtzman found:

3. The fact that project cancellation rates are highest in low-income areas is entirely explained by their higher resistance rates.

“Projects were six times more likely to be canceled in communities that resisted than in communities that chose to comply,” Holtzman points out. He added: "The increase in cancellation rates in low-income areas is entirely due to the high protest rates in these communities. Therefore, if the project proposals continue to be promoted in these areas, it may arouse more public outrage, trigger stronger resistance, and further drive the cancellation rate of the project."

I hope this data helps shatter the arrogance that the data center boycott movement is led by wealthy NIMBYs. In fact, the vast majority of those who stood up to resist were working-class residents and communities. I also hope these conclusions provide a powerful weapon for cities, residents, and organizers struggling with data center development.

My sincere thanks again to Holtzman for allowing me to publish these findings on my blog. For those interested in studying or further examining his data, he has hosted the entire code repository on GitHub.

The United States as a whole has become the country most resistant to new data centers

The data comes from the research organization Public First (thanks to WIRED reporter Molly Taft for sharing): How did the United States, as the center of the AI boom, become a stumbling block to it? Our investigation offers several explanations.

–Informed Objection

The public understands more than ever what AI is and does, and what data centers are and do. When we surveyed AI 5 years ago, it was a fringe interest at best. We are now seeing significant growth in public awareness and understanding, and greater maturity in the use of tools, particularly in the 25-44 age group. Our analysis of who understands AI needs to shift from “who turns on large language models” to “who uses large language models in complex, integrated ways.”

Our survey shows that the U.S. ranks in the middle of the pack in terms of self-professed knowledge of data centers, higher than other “developed” markets. This is not surprising considering how widespread data center construction is in the United States.

And this “informed opposition” leads it to hate data centers more than any other country surveyed, interesting!

Note 1:

According to Holzman: I used 5-year American Community Survey data from the years 2020-2024, so the income numbers are generally a little lower than you might expect. I need to do this to get data at the census tract level; so for the national median, I stick with the same data set.