New York Times AI Poll Finds Data Centers Unpopular, but Neither Party Owns the Issue

The New York Times AI poll found 61 percent of likely voters oppose building data centers for AI, yet neither party holds a convincing political advantage.

That combination matters more than the topline alone. Data center opposition now crosses party lines, while voters remain almost evenly divided over which party they trust on AI.

The result creates an unusual midterm dynamic. Candidates can criticize AI infrastructure without offering a settled national policy, and voters can distrust the construction boom without switching party loyalties.

The national polling also shows why the issue is difficult to capture with a conventional partisan message. Republicans are more open to data centers than Democrats, but nearly half still oppose them. Democrats express stronger opposition, yet they have not established a decisive advantage on AI governance.

The real contest is therefore not Republicans against Democrats. It is the technology sector's promise of national progress against voters' experience of local costs, uncertain benefits, and limited control.

The New York Times AI Poll Reveals Broad Opposition

The poll turns dissatisfaction with AI infrastructure into a measurable national political problem.

The New York Times and Siena University surveyed 1,503 likely voters nationwide from September 8 through September 13, 2026. Sixty-one percent opposed constructing data centers to support AI technology, while 34 percent supported it.

The partisan breakdown shows a difference in intensity, but not a clean division. Seventy-five percent of Democrats opposed construction, compared with 60 percent of independents and 47 percent of Republicans.

Republicans were the only group in which support narrowly exceeded opposition. Forty-nine percent supported construction, while 47 percent opposed it. That four-point difference hardly amounts to an enthusiastic constituency.

Age produced an even sharper pattern. Nearly 80 percent of voters under 30 opposed construction. Opposition declined with age, but 53 percent of voters over 65 still opposed it.

That result reverses a familiar assumption about technology adoption. Younger people often use AI tools more frequently, yet they are also the most resistant to the physical infrastructure supporting those tools.

The question covered data center construction across the United States, not only projects proposed in a respondent's immediate community. This distinction makes the opposition especially notable.

Local polling often produces a familiar response: people support development in theory but reject a project near their homes. The New York Times AI poll suggests skepticism has moved beyond that neighborhood boundary.

Still, the result does not establish that most voters want a complete prohibition. Among respondents who opposed data centers, 56 percent favored limits rather than an outright ban.

That preference introduces an important policy distinction. Many voters appear to reject uncontrolled construction, not every facility under every possible condition.

The reasons also varied. Climate-related concerns were the most common, cited by nearly one-third of opponents. Eighteen percent cited a general distrust of AI rather than one specific infrastructure cost.

Democrats were more likely to mention environmental concerns. Republicans were more likely to express broader suspicion toward AI itself.

These differences give candidates multiple ways to describe the same opposition. One campaign can focus on utility bills and water. Another can emphasize farmland, emissions, employment, or mistrust of technology companies.

Yet the shared conclusion remains consistent: unconditional approval for more AI infrastructure has become politically difficult to defend.

This AI data center opposition is no longer based on one controversial development. It now appears in national polling, local fights, campaign strategy, and debates over electricity regulation.

Voters Dislike Data Centers but Still Do Not Prioritize AI

AI has become broadly unpopular without becoming a decisive voting issue.

Fewer than 1 percent of respondents identified AI or data centers as the most important issue determining their November vote. The economy remained the leading priority.

That figure places a major limit on interpretations of the poll. Candidates face political risk when they defend data centers, but opposition alone is unlikely to decide most congressional races.

Instead, AI infrastructure connects with issues that voters already rank highly. Electricity prices affect household budgets. Water use becomes a local resource dispute. Tax incentives raise questions about public spending.

A data center can therefore become politically important without voters describing AI as their primary concern. The project acts as a concrete example of broader anxiety over affordability, development, and government responsiveness.

This helps explain why campaign messages increasingly focus on who pays. Voters may not follow debates about model training or computing capacity, but they understand a higher utility bill.

Independent polling has found similar concerns. An Annenberg Public Policy Center survey reported that local opposition rose from 49 percent to 61 percent between early spring and midsummer.

That survey covered 1,320 adult citizens and carried a margin of error of 3.5 percentage points. It found majorities of Democrats, Republicans, and independents opposing new data centers in their areas.

The movement over time may be more important than any single result. Opposition increased by 12 points within approximately four months, the largest change among the survey's AI questions.

Views about AI's overall effect did not shift significantly during that period. Thirty-nine percent expected a negative national impact over the next decade, while 18 percent expected a positive one.

This separation suggests that data centers are developing their own political identity. Voters do not need to reject every AI application to oppose the infrastructure used to develop and operate it.

Even frequent AI users showed little additional enthusiasm for local construction. The Annenberg survey found opposition at 60 percent among heavy users, compared with 64 percent among nonusers.

Familiarity with chatbots or workplace AI may improve attitudes toward the software. It does not automatically reduce concerns about power consumption, privacy, employment, or construction near homes.

A separate University of Massachusetts Amherst survey found an even wider gap. Its community polling showed 65 percent opposed a local AI data center, while only 11 percent supported one.

Among Republicans, 52 percent opposed local construction and 18 percent supported it. Democrats opposed it by 76 percent to 7 percent.

Respondents most frequently cited natural resources, land use, local disruption, utility costs, and distrust of the AI industry. Those categories overlap, but each supports a different policy response.

The findings weaken the idea that opposition is simply fear of unfamiliar technology. Voters are evaluating visible tradeoffs associated with facilities that consume land, electricity, and water.

For developers, this creates a communications problem that cannot be solved by promoting AI products. A useful chatbot does not answer who finances a new transmission line.

For politicians, it creates a governing problem. Candidates must translate general opposition into standards that can distinguish an acceptable project from an unacceptable one.

Neither Party Has an Edge on AI

Voter frustration is widespread, but it has not produced clear ownership of the issue.

Asked which party they trusted more to handle AI, 42 percent chose Republicans and 40 percent chose Democrats. That two-point difference fell within the poll's margin of error.

The Republican advantage also contrasts with the poll's other issue results. Democrats led on the economy, cost of living, immigration, foreign wars, and the federal deficit.

An issue comparison found notable demographic divisions beneath the narrow national result. Republicans led on AI among voters aged 45 to 64 but trailed substantially among voters under 30.

Republicans also performed better with men than with women. These gaps suggest neither party has assembled a broad coalition around its approach to AI.

The answer to which party handles AI therefore depends heavily on who is being asked. It also depends on whether the voter is thinking about innovation, safety, employment, energy, or local construction.

Republican leaders have often emphasized rapid development, reduced regulation, and competition with China. President Donald Trump has defended data centers as engines of growth and national strength.

House Speaker Mike Johnson has warned against broad restrictions that might weaken American technology development. That message aligns with companies seeking faster permits and greater power capacity.

However, some Republican candidates are moving toward guardrails. Michigan Senate nominee Mike Rogers has called for stronger AI regulation, placing himself at odds with the administration's more permissive direction.

Democrats have their own internal divisions. Progressive figures, including Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez, support much stronger restrictions on advanced AI development.

Other Democrats focus more narrowly on electricity bills, environmental reviews, labor protections, or local authority. Many have stopped short of supporting a general pause.

Those divisions prevent either party from presenting one recognizable AI program. Voters may hear different positions from presidential leaders, congressional candidates, governors, and local officials wearing the same party label.

The problem becomes especially visible with data centers. National leaders can describe the facilities as strategic assets, while candidates near disputed projects treat them as political liabilities.

This fragmented positioning explains the poll's central contradiction. Republicans receive a slight trust advantage even though their national leader strongly supports infrastructure that most voters oppose.

Democrats express greater skepticism toward data centers, but voters have not rewarded them with broad confidence on AI. Opposition is not the same as policy credibility.

Previous polling reached a similar conclusion. Ipsos found that Republicans were more supportive of construction, but still more likely to oppose it than support it.

It also found widespread uncertainty about each party's approach. At least half of both Democratic and Republican respondents were unsure which party's AI policies they preferred.

The New York Times AI poll updates that picture closer to the midterms. Voters recognize the conflict, but they still cannot identify one party as the reliable manager of it.

That leaves room for individual candidates to create their own positions. It also makes those positions vulnerable to contradiction from national party leaders.

A candidate can promise local limits while belonging to a party that promotes faster construction. Another can campaign against projects while accepting that AI services require expanding computing capacity somewhere.

Neither side has resolved that tension. Until one does, which party handles AI will remain a question without a stable answer.

The Real Divide Is National Ambition Versus Local Cost

The strongest political conflict pits the national promise of AI leadership against the local burden of building its infrastructure.

Supporters present data centers as necessary foundations for economic growth, scientific research, national security, and competition with China. The argument operates at national scale.

Opponents experience the projects differently. They see rezoning requests, industrial buildings, transmission infrastructure, water demand, backup generators, and negotiations over tax treatment.

Both perspectives can describe real aspects of the same project. The political imbalance comes from how costs and benefits are distributed.

A company can promise wider economic gains from AI while placing one facility's resource demands in a particular county. Residents must judge the immediate proposal, not an abstract national strategy.

Jobs frequently anchor the case for approval. Construction can create substantial short-term employment, while permanent operations require technical, security, and maintenance workers.

However, residents also ask how many lasting jobs will remain after construction. They want to know whether those roles will go to local workers and justify the public concessions involved.

The same mismatch appears with electricity. Developers need reliable power at enormous scale, but neighboring households care about whether new demand changes their monthly bills.

Water creates another point of conflict. Some data centers require significant cooling resources, although designs and local conditions vary widely.

Broad claims about average water consumption cannot resolve a site-specific dispute. Communities need project-level information about cooling systems, drought conditions, and competing users.

The bipartisan nature of the backlash follows from this local focus. Conservative landowners can oppose government-backed industrial development, while environmental groups object to energy and water demands.

Labor organizations may support construction jobs even when Democratic activists oppose the same project. Business groups may align with Republican leaders while local Republican voters resist development.

The Associated Press documented this bipartisan backlash in communities where farmers and environmental advocates found common ground.

These coalitions do not require agreement about AI, climate policy, or national politics. They only require agreement that one proposed development creates unacceptable local risks.

That structure makes AI data center opposition hard for national parties to absorb. Conventional political messages usually depend on stable ideological groups.

Data center disputes produce temporary alliances organized around geography, infrastructure, and household costs. Those alliances can dissolve when the location or project terms change.

The polling question also affects the apparent level of opposition. Asking about construction nationwide produces one result. Asking about a facility in a respondent's community produces another.

The New York Times AI poll found 34 percent supporting national construction. The UMass survey found only 11 percent supporting construction in the respondent's community.

Those surveys used different samples and wording, so their results should not be treated as a direct trend line. Together, they still illustrate the importance of proximity.

A national party can endorse data centers in principle. A local candidate must answer whether a specific developer receives permits, grid access, water, or tax treatment.

The technology industry's most difficult task is therefore not convincing people that AI has potential. It is proving that individual projects allocate their costs fairly.

That requires enforceable commitments, transparent resource estimates, and credible oversight. General promises about innovation cannot substitute for those details.

Conditional Approval Is Becoming the Political Middle Ground

Voters appear more receptive to enforceable conditions than to either unrestricted development or a total ban.

The New York Times and Siena finding supports this interpretation. Most opponents preferred limits, while a smaller share wanted data centers prohibited entirely.

Republican strategists have already recognized the distinction. Trump pollster Tony Fabrizio warned candidates that unconditional support for more AI data centers had become a losing position.

His campaign guidance advised candidates to validate concerns about utility bills, grid pressure, and water before discussing potential benefits.

The guidance proposed a clear standard. Developers should supply their own power needs, protect consumers from higher bills, and safeguard local water and natural resources.

Fabrizio's polling found 61 percent of registered voters favored construction with conditions. Twenty-eight percent favored prohibiting data centers altogether.

Without conditions, voters opposed additional construction by 65 percent to 24 percent. These findings came from separate polling and should not be merged numerically with the Times and Siena results.

The broader pattern is nevertheless consistent. Opposition declines when voters receive assurances addressing specific household and community costs.

This middle ground creates a test for both parties. It is easy to promise guardrails, but harder to define who enforces them and what happens when projections prove wrong.

A requirement that developers protect ratepayers needs a measurable baseline. Regulators must determine which grid upgrades belong to a project and which benefit the wider system.

Promises about water protection also require disclosure. Officials need reliable estimates, monitoring rules, and consequences for exceeding agreed limits.

Local governments face their own capacity problem. Smaller jurisdictions may negotiate with global technology companies and utilities using limited technical staff.

That imbalance can increase public distrust even when a proposed agreement contains safeguards. Residents need enough time and information to evaluate the terms independently.

Candidates also risk overstating what conditional approval can accomplish. A project can meet formal rules while still changing land use, traffic, or the surrounding economy.

Conversely, not every feared effect will materialize at every site. Data centers differ in size, power source, cooling design, ownership, and workload.

This variation makes a blanket national judgment difficult. It also strengthens the case for project-specific evidence instead of generic industry assurances.

The skeptical view is that conditional approval may function mainly as campaign language. Politicians can endorse strict standards without rejecting a politically connected development.

That concern will grow if rules remain voluntary. It will weaken if utilities, regulators, and local governments publish enforceable agreements before construction begins.

Technology companies face the same credibility test. They can announce sustainability goals, but communities will judge contracts, permits, and utility filings.

The industry's previous message emphasized scale, investment, and geopolitical competition. The emerging political standard emphasizes responsibility for direct costs.

That is a meaningful reversal. Developers once sought public support by describing what their projects would add. Now they must demonstrate what residents will not be forced to absorb.

Neither party has fully converted that standard into national legislation. Until they do, data center policy will continue to develop through state rules and local negotiations.

Three Signals Will Show Whether the Backlash Reshapes Policy

The next stage will be measured through enforceable utility rules, midterm campaign behavior, and approval outcomes for contested projects.

The first signal is action on electricity costs. Lawmakers are considering ways to prevent data center demand from increasing household power bills.

The details matter more than the political announcement. A meaningful policy would specify how utilities allocate generation, transmission, and grid-upgrade expenses.

If developers receive separate rates or must finance dedicated capacity, the conditional-development position will gain credibility. If households continue absorbing disputed costs, opposition will likely intensify.

The second signal is how candidates describe AI data centers during the final weeks before the midterms. Watch whether they endorse construction broadly or attach measurable conditions.

Republicans face a visible conflict between Trump's support for rapid expansion and warnings from their own campaign advisers. Their candidates must decide which message fits their districts.

Democrats face a different challenge. Criticizing local costs can attract voters, but the party still needs a coherent position on AI development and national competitiveness.

If candidates in both parties converge on similar conditions, the issue may remain bipartisan rather than becoming a conventional partisan divide.

If one party pairs those conditions with credible legislation, it can begin answering which party handles AI. Polling should then show movement beyond the current 42-to-40 split.

The third signal is what happens to contested projects. Permit denials, moratoriums, ballot measures, and renegotiated development agreements will reveal whether polling becomes policy.

A project that proceeds after accepting stronger power and water protections would support the conditional-approval model. A wave of cancellations would indicate a deeper rejection.

Approval without meaningful concessions would test whether public opposition carries electoral consequences. Local races may provide that evidence sooner than national polling.

These signals also matter to developers and enterprise AI customers. Delays in new computing capacity can affect expansion schedules, infrastructure planning, and the availability of AI services.

Knowledge workers do not need to follow every zoning hearing. They should understand that the software they use depends on physical systems facing growing political scrutiny.

The New York Times AI poll does not show that Americans have rejected AI outright. It shows that the infrastructure bargain has lost public confidence.

Neither Republicans nor Democrats currently own the solution. Both can describe the anxiety, and both contain factions pulling toward faster development or tighter limits.

The opening now belongs to whoever can convert public concerns into rules that voters consider enforceable. That means answering who pays, who benefits, and who can stop a harmful project.

Over the next three months, watch utility policy, candidate language, and contested permits. Together, those signals will show whether AI data center opposition remains diffuse anger or becomes a durable governing agenda.

The practical question is no longer whether the United States will need more computing infrastructure. It is whether elected officials can establish acceptable terms before public resistance hardens further.