By This Hour Technology Desk
Public resistance to building data centers for artificial intelligence appears to be broad, crossing some political lines while stopping short of becoming a central voting issue. A reported New York Times-Siena University survey of likely voters found that 61% opposed construction of data centers intended to power AI technology. Only 14% said they strongly supported it.
The finding matters because the argument over AI is no longer confined to the software’s uses, its safety, or its effects on work. Data centers put the physical requirements of the technology in view: new facilities, local land-use decisions, demands on water and environmental resources, and questions about who bears the costs. The survey results suggest those concerns may be giving people who otherwise disagree politically a reason to resist further construction.
The available reporting describes a sharp overall imbalance in sentiment, but not a simple partisan alignment. It also indicates that voters who object do not all favor the same response. Many prefer limits on construction, while a substantial minority favor an outright ban. That distinction could become important wherever officials are asked to choose between approving projects, imposing conditions, or blocking them.
Opposition is broad, but the partisan split is uneven
The New York Times-Siena poll, as described in the available report, surveyed 1,503 likely voters in early September. Its central question asked respondents whether they supported or opposed construction of data centers to power AI. The 61% opposition figure is the broadest measure in the results supplied here, and the 14% figure refers specifically to people who said they strongly supported construction. The material does not provide a full breakdown of every response category, so the gap should not be treated as a complete account of public opinion.
Even so, the reported subgroup results complicate any assumption that data-center opposition belongs neatly to one party. Among people who voted for Kamala Harris in 2024, 74% were reported to oppose construction. Among people who did not vote, opposition was reported at 67%. Voters who backed Donald Trump in 2024 were much more evenly divided: 49% supported construction and 45% opposed it.
That divide gives neither major party an uncomplicated advantage. Opposition among Harris voters is pronounced, while Trump voters are close to split rather than uniformly supportive. The nonvoter figure adds another wrinkle. A large reported share of that group opposed construction, but the polling material supplied does not explain whether this reflects concern about particular projects, broader views of AI, or other characteristics of respondents who did not vote in 2024.
Nor does an opinion on data centers automatically reveal an opinion about AI in every form. The survey question, as summarized, focused on facilities built to power AI technology. A respondent may oppose siting a large facility near home while using AI products, supporting AI research, or accepting data centers in other places. Conversely, a person may object to AI itself and express that view through opposition to its infrastructure. The poll’s reported answers identify a shared objection to construction, not a single political or technological worldview.
Water, environmental effects and local impacts shape the objections
The reasons offered by opponents show why the issue can unite otherwise different constituencies. Environmental concerns or water use were the most frequently reported reason, cited by 32% of respondents who opposed data centers. Effects on local communities followed at 21%. General distrust or dislike of AI accounted for 18%, while regulation concerns and economic concerns were cited by 10% and 9%, respectively.
These categories are distinct, but they can overlap in practice. Concern about water or the environment addresses the resource implications of a facility. Concern about a local community addresses how a project is experienced in a particular place. Distrust of AI directs attention to the technology the facilities are meant to serve. Regulation and economics turn toward public oversight and perceived distribution of costs or benefits. The ranking indicates that resistance is not based only on a generalized fear of unfamiliar technology.
Environmental and water concerns led the list, but the numbers supplied do not show what respondents meant by those terms, whether they had a particular project in mind, or how they weighed them against possible benefits. The local-community category is similarly broad. It may capture worries about siting and neighborhood consequences, but the available record does not specify which effects respondents identified. Those gaps matter when translating national survey findings into decisions about individual proposals.
Still, the ordering of concerns gives a clearer picture of what a public case for new construction would need to address. A message focused solely on AI’s capabilities would not directly answer objections centered on resources or local consequences. Nor would a generic promise of economic benefit settle concerns about environmental use, community impact, or confidence in oversight. The poll does not prescribe a policy, but it suggests that the public debate is likely to turn on concrete conditions rather than technological ambition alone.
Related reporting has raised questions about environmental risks associated with the wider data-center push. Former EPA officials warn data-center push could raise pollution risks examines competing claims about federal environmental policy. That separate account does not validate the polling figures, nor does it establish that any particular facility will produce the effects feared by respondents. It does, however, illustrate why environmental oversight has become part of the wider argument around AI infrastructure.
Most opponents favor constraints rather than a universal prohibition
The reported survey also separates opposition to construction from support for an absolute stop. Among respondents who opposed data centers, 56% favored limits and 38% favored a total ban. The difference is consequential. A preference for limits leaves room for rules, conditions, and selective approvals; a preference for a ban does not. The survey therefore points to a public that is resistant overall but not uniform in its desired remedy.
That distinction may shape how elected officials interpret the result. Treating every opponent as a supporter of a nationwide or universal ban would go beyond the figures. Equally, treating support for restrictions as consent for unrestricted expansion would misread the reported majority within the opposition group. The available results support a narrower conclusion: many people who object to new AI data centers want boundaries placed on construction, while a smaller but still sizeable share want it stopped altogether.
The reporting does not specify what kinds of limits respondents favored. It does not say whether they meant restrictions connected to water, environmental considerations, community effects, AI governance, economic terms, or another concern. It also does not show how those preferences would change if people were presented with different safeguards. Any claim that the poll settles which regulatory response voters want would therefore exceed the evidence supplied.
The survey’s political results carry a similar limitation. The close division among Trump voters could make the issue contested within that coalition, but it does not establish how candidates will address it. The larger opposition reported among Harris voters and nonvoters does not establish that those groups will vote on the issue. Attitudes can be strong without determining electoral behavior, particularly when voters rank other subjects higher.
A strong view that may not move the midterm vote
That is the central tension in the reported findings. Despite the high level of opposition to data-center construction, AI and data centers were said to rank very low among issues voters considered most important for their midterm choice. For most demographic groups, they accounted for less than 1% of responses. Voters aged 18 to 29 were an exception, with 3% selecting the issue.
The gap between opposition and salience is significant. It suggests that an issue can draw a negative reaction when people are asked about it directly, yet remain behind other concerns when they decide how to vote. The supplied reporting identifies the economy, foreign policy, inflation, and opposition to Trump or Republicans among the more common answers in the broader issue-ranking question. It does not provide enough information to measure how durable the low ranking is or whether it could change around a local construction dispute.
For political parties, that combination offers limited grounds for confidence. Voters may not treat AI infrastructure as a top national electoral test, but local projects can still make the issue immediate for affected communities. The poll does not measure local contests, project-specific campaigns, or whether views change once a proposal is announced. Its value is in showing a broad national mood toward construction, not in predicting the outcome of a particular approval process or election.
Two other polling references in the available report point in the same general direction, though they should be handled carefully because their underlying materials were not supplied for review. An NBC News poll in March was described as placing AI below ICE in public polling. A Gallup poll in May was described as finding that roughly seven in ten Americans opposed local construction of AI data centers. These references are consistent with the Times-Siena result, but differences in wording, timing, population and methodology can materially affect polling results.
The evidence available here supports a cautious conclusion rather than the sweeping claim that AI and data centers are unpopular in every poll. It indicates that several reported polls have found substantial resistance, with the Times-Siena survey offering the most detailed figures in the supplied material. It does not establish unanimity, identify all surveys on the subject, or prove that every form of AI infrastructure faces the same reaction.
This report has not been independently corroborated. The figures and descriptions in this article are drawn from a single secondary report summarizing the polls, and the underlying questionnaires, methodology, full crosstabs and original polling releases were not available in the supplied record. Those omissions limit the ability to assess question wording, sampling, margins of error, subgroup reliability and whether apparent differences across polls reflect genuine shifts in opinion.
Reporting notes
What is confirmed: The supplied report gives figures for overall views, voter groups, preferred responses and reasons for opposition.
Why this matters: The findings connect AI expansion to concerns about resources, local effects and public oversight.
What remains unclear: Original poll materials and the methods behind other referenced surveys were not available for review. This report is based on one source and has not been independently corroborated.
