July 21, 2026

SHARE |

A Year of Monitoring: Anti-Data Center Narratives and Threats

(AP Photo/George Walker IV)

The state of New York has established the United States’ first statewide moratorium on hyperscale data centers, framed as a bid to keep residents’ electricity bills down and protect environmental resources. Legislative proposals elsewhere similarly seek to slow the buildup of computational infrastructure that has become an important part of AI research, development, and deployment. Michigan is attempting to pass a similar ban to New York on data center construction. In Monterey Park, California, residents voted to permanently ban data centers in the city. There are also several efforts in Washington, D.C. to establish federal rules and guidelines concerning AI data centers. One act looks to amend an existing Federal Power Act to give the Federal Energy Regulatory Commission authority to regulate and establish firm guidelines on how data centers or other large-load electricity users connect to an interstate electricity grid. Another proposed bill looks to curb the rising energy bills many U.S. citizens are noticing by requiring state utilities to consider creating a "large load standard" that shifts the cost of grid upgrades needed to power data centers onto the facilities themselves rather than everyday ratepayers. Outside the U.S., the city of Amsterdam has banned the construction or expansion of new data centers until at least 2030, and Denmark's legislature has introduced a bill to reduce data centers' priority access to the power grid. Germany has restricted new hyperscale facilities over 70 megawatts to just two designated locations nationally, citing grid congestion.

Legislative action in the U.S. is a direct response to the growing unpopularity surrounding AI data centers. A recent Gallup poll found that “Seven in 10 Americans oppose local construction of AI data Centers,” as opposition to data centers continues to grow as a hot-button political issue. Traveling around the United States by car, it is not unusual to see front lawns adorned with anti-AI data center signs, an anecdotal yet still pertinent observation. Opposition to data centers also seems to be, at least at the citizen level, a bipartisan issue, with protests and demonstrations in both red and blue states.

In parallel with governmental responses and peaceful activism against the rapid construction of AI infrastructure, there has been a spike in incidents indicating escalation in the anti-data center movement. Last week, the activist group Extinction Rebellion in the Netherlands hurled balloons containing hydrogen peroxide and acetic acid at a Microsoft data center, citing concerns about the undemocratic power of a small group of the very wealthiest and about electricity use. In May, in Michigan, the Saline Township Treasurer announced her resignation, citing death threats, as a contested data center was being constructed. In June, the disrupted plot at the UFC Freedom 250 event at the White House, which aimed to kill senior government officials, was motivated by a range of grievances, including concerns about data centers’ water consumption.

The Soufan Center has been monitoring anti-data center conversations and sentiment across social media platforms (primarily X and Reddit), both through automated scraping and manual data collection, to chart the volume and narratives of explicit threats as well as mainstream opposition. In the July 2025 to July 2026 window, our dataset, which focuses on data center discourse (rather than AI discourse broadly), shows a high baseline of volume with a notable surge beginning in April 2026 that has elevated that baseline. The pattern is largely event-driven, but it appears that a narrative environment has established that each new incident related to data center construction, or community pushback against it, generates much greater secondary amplification than it did earlier in the year — meaning once a story breaks, they compound onto each other, feeding back into the information environment and causing secondhand outrage. This type of amplification can appeal to people from all walks of life and from various political and socioeconomic backgrounds, making it potentially all the more dangerous, as public anger continues to grow.

In terms of mainstream narratives, with the Boolean query used, data center opposition primarily revolved around energy and grid strain on the one hand, and water use and quality on the other. The first narrative centers on the impact of data centers on grid capacity, power demand, and electricity costs for communities living near data centers, as well as who bears those costs, with broad criticism of the lack of AI companies stepping in for the impact on electricity bills. Water use narratives were the other main area of mainstream opposition to data centers, which focused on concerns about water usage by data centers as well as fears of water pollution. Narratives around data centers’ use of zoning laws and eminent domain — a U.S. legal clause under the Fifth Amendment which allows the government to seize private property for public use even at the objection of the property owner —  are less continuous and appear alongside headlines about forced displacement and property seizures for data-center-related infrastructure. Quality-of-life issues related to noise and visual pollution were another clear bucket of opposition but appeared less in the dataset, likely primarily emanating from people directly confronted with the issue. While the dataset collection focused on data center discourse, broad AI concerns were also identified, primarily fears about its impact on jobs and the economy, as well as worldwide environmental impact, and issues about privacy and surveillance.

The peaks in volume of general data center opposition content are largely event-driven. For example, one big amplification cluster in this dataset is a late-May 2026 storyline tying Meta’s data center construction to drinking water degradation in Georgia. On May 18, More Perfect Union, a progressive non-profit newsroom, took U.S. Congresswoman Alexandria Ocasio-Cortez to Georgia in a now-viral video to meet with Georgia residents who had been claiming their water was being contaminated by the construction of a Meta-owned data center. Days later, during a congressional hearing, Representative Cortez pressed a member of the U.S. Environmental Protection Agency (EPA) on this issue, giving the story further attention when she brought a jar of brown water from Georgia. Other clusters of negative sentiment come from viral stories of local pushback against data centers, such as a Kentucky farmer family in Mason County that turned down a $26 million offer to convert part of their farm into a data center. Another cluster of negative posts comes from a family in Georgia who are being forced to sell their home under threat of eminent domain for a new transmission line that utility company Georgia Power plans to build. Only 20-30 percent of power from the new line will serve residential communities, while 70-80 percent will go towards data centers.

It is no surprise that these issues would serve as important amplification clusters in the dataset. The narratives represented — water contamination (which exists as part of the larger grievance of environmental degradation) and property rights — are engaging but are not the primary focus of U.S. lawmakers looking to regulate the construction of data centers. Most legislation focuses on energy grid use and the rising cost of utilities, which do remain of significant concern to U.S. citizens. However, the more visceral nature of other issues, such as undrinkable water or the loss of one’s home, is particularly suited to social media engagement. These narratives carry the greatest potential to radicalize individuals, driving acts ranging from vandalism against data centers to violence against corporate and political figures.

Compared with general anti-data-center commentary, threat content (credible and non-credible) is far less voluminous and appears in intermittent bursts that subsequently dissipate. Throughout the observation window, our research identified hundreds of posts and comments that included threat language across a spectrum — from harassment to intimidation to actual threats of direct violence. The most prevalent category of direct violent threats was directed at infrastructure rather than individuals. Many of these posts do not explicitly use threatening language and seek to evade content moderation through memes, emojis, and deliberate misspellings of words that may be filtered out, such as arson and bombs. Some posts also threaten council members and planning commissions, as well as the companies involved in constructing these data centers, but these remain a small minority. Less direct threats that appear more rhetorical than credible share a recurring linguistic pattern in which explicit language like “fire,” “arson,” or “kill” is used with a general symbolic target (“a data center,” “everyone involved”) rather than a specific location, person, or data center. As with verbs such as “kirked” and “luigi-ed” that have come to mean ‘to be killed’ in reference to the late political pundit Charlie Kirk and United Healthcare assassin Luigi Mangione, respectively, these have been used predominantly in ways that appear largely non-credible. GIFs of fire, guillotines, and bombs have been observed under announcements of data centers without a clear sign of intent. While threat content represents only a small fraction of general anti-data-center commentary, these issues are clearly ripe for engagement and are likely to remain a contributing factor in the broader radicalization landscape.