Version 13

Community opposition to AI data centers has blocked 75 U.S. projects worth approximately $130 billion in Q1 2026 [1], with over 300 state and local bans or moratoria enacted since 2023 [2]. Seattle passed a one-year moratorium [6] and Oklahoma City approved its own [7], while New York's legislature passed the country's first state moratorium awaiting the governor's signature [11][12]. A Gallup poll shows 71% of Americans oppose local data center construction, with bipartisan majorities of Democrats, independents, and Republicans all opposed, driven primarily by resource concerns [2]. Tech voices John Carmack and Marc Andreessen counter that opposition is misinformation-driven [21][22], while NVIDIA argues its Rubin architecture eliminates water consumption in next-generation facilities [24].

The bipartisan nature of public opposition — with majorities of all three partisan groups against local construction [2] — removes the standard partisan segmentation that industry lobbying typically exploits. Municipal and state restrictions continue to multiply faster than federal frameworks can override them, creating a patchwork of restrictions that complicates national-scale AI infrastructure buildout.

AI data centers rely on evaporative cooling that consumes fresh water at scale, and communities near planned facilities have built a coordinated set of tools to block or reshape projects: water rights protests, local zoning, ballot initiatives, and state legislation. Data Center Watch reported that community protests blocked or delayed at least 75 U.S. projects worth approximately $130 billion in Q1 2026, the highest quarterly total since tracking began in 2023, with active opposition groups more than doubling to 833 across 49 states [1]. Over 300 state and local data center bans or moratoria have been enacted since 2023, with resistance strongest in the Midwest and South [2]. A Gallup poll found 71% of Americans oppose local data center construction, with opposition crossing partisan lines — majorities of Democrats, independents, and Republicans all opposed — driven primarily by resource concerns: 18% of opponents cite water use and 18% cite electricity demand [2]. Food and Water Watch published a community organizing guide for opposing data centers [3], formalizing the opposition's infrastructure.

Permanent bans are in place in Monterey Park, California (86% voter support) [4] and St. Charles, Missouri [5]. Seattle's City Council passed a one-year moratorium [6], and Oklahoma City's Council approved a moratorium on new data centers [7]. The Utah Stratos project was cut by 50% after water rights protests [8][9]; Avista utility paused a 500-megawatt MOU in Spokane County after more than 5,000 community comments [10]. At the state level, New York's legislature passed a data center moratorium — the first state legislature to do so — awaiting the governor's signature [11][12]. South Carolina is tracking H5526 [13]; Michigan lawmakers proposed a one-year pause [14]; Illinois paused new data center tax incentives [15]; Ohio activists placed a construction prohibition on the 2026 state ballot [16]; and a federal Senate bill, S.4214 (AI Data Center Moratorium Act), has been introduced in the 119th Congress [17].

The federal government has moved in the opposite direction. FERC approved faster grid connections for AI data centers [18], and the Trump administration has reportedly considered Cold War-era emergency authorities to direct resources toward AI infrastructure [19]. Michigan Governor Whitmer preserved state tax incentives and appeared with Sam Altman at a Michigan facility [20]. In late June 2026, prominent tech voices entered the public debate with arguments that go beyond aggregate-scale minimization. John Carmack argued that anti-datacenter activism resembles the anti-nuclear movement — driven by 'vibes' rather than evidence — and that data center demand reflects legitimate market value for AI [21]. Marc Andreessen dismissed water concerns as a 'completely fake meme' [22]. This framing differs from the argument that aggregate AI water withdrawals are a small fraction of total U.S. water use [23]; Carmack and Andreessen assert the underlying concern is fabricated, not merely overstated in scale.

NVIDIA announced in June 2026 that its Rubin platform achieves 100% liquid cooling with no fans, and that its closed-loop system reduces facility water consumption to near zero in suitable climates [24][25]. The company states that a 50-megawatt facility using this architecture saves more than $4 million annually in cooling costs. This shifts the industry argument from minimizing aggregate water use to claiming the problem can be engineered away in next-generation infrastructure — a claim whose practical effect on ongoing disputes depends on how quickly the Rubin architecture deploys and whether regulators accept future technology commitments as grounds for approving projects that currently use evaporative cooling. A UN-affiliated institute projects AI data center water use will reach 9.3 trillion litres annually by 2030 [26], a trajectory that makes current aggregate comparisons less informative over time.