This is a seventeen-slide polling deck produced by the research and strategy consultancy Public First and released in August 2026 by the Information Technology and Innovation Foundation, at a webinar titled "Can the United States Win the AI Race if Americans Oppose Data Centers?" ITIF president Daniel Castro framed the findings as a risk that opposition to data centers becomes opposition to AI. The deck is commissioned research published by an organization with a stated pro-technology position, and its findings should be read as such.
Two instruments, not one
The deck draws on two separate polls, and their attribution matters because they ask different questions.
Public First AI Global 2026 is a fifteen-country survey. It supplies the cross-country comparison, the partisan-gap comparison, and the year-on-year trend in general attitudes to AI. Its question about data centers is national in scope — "Should your country build more or fewer data centers?"
The POLITICO Poll by Public First supplies the January-to-July 2026 trend line, the belief-change data and the party-ownership data. Its data-center question is local and specific — "Would you support or oppose the following being built in your local area (within 3 miles of where you live)?"
Ten of the seventeen slides carry no source line, so which instrument produced them is not established by the document. More consequentially, the deck states no field dates, no sample sizes, no margins of error and no fieldwork provider for either instrument. The only methodological note in the document is "National totals, weighted, 2026" on the cross-country slide. Findings drawn from it therefore cannot be compared with confidence against polls whose methodology is published, and its confidence rating here is medium for that reason rather than because the figures are in doubt.
Data centers against other development
The deck's opening finding sets a data center against nine other things that might be built locally. Asked "Would you support or oppose the following being built in your community?", respondents give a new data center a net score of −20, the only negative score in the set. New housing scores +43, a new road or highway +30, a new solar farm +25, a warehouse and a wind farm +21 each, a train station +18, a shopping mall +16, an office complex +10, and a new power plant or other electric infrastructure +2.
The gap holds under a forced choice. Asked to imagine undeveloped, unused land near where they live with "no other viable options for what to do with it," respondents prefer to leave the land undeveloped by a net of −15. The margin widens by locality (urban −8, suburbs −16, small town −25, rural −24), by partisanship (would vote Republican +1, would vote Democrat −25), and above all by trust in technology companies: those who trust them completely or somewhat are at +16, those who trust them a little at −29, and those who do not trust them at all at −48.
Distance is the strongest single moderator. Asked about data centers "in the United States," respondents are positive about data centers in a different state (+20), data centers used for online banking, video streaming and photo storage (+18), data centers anywhere in the world (+16), data centers storing government data (+11) and data centers within their own state (+7). They turn negative on data centers used to train AI models (−2) and sharply negative on data centers "within a 10 minute drive of where you live" (−19).
The cross-country comparison
Asked "Should your country build more or fewer data centers?", the United States ranks fifteenth of fifteen countries at +2 net (24% build fewer, 26% build more). The full ranking is Nigeria +68, Vietnam +59, India +58, Indonesia +48, Brazil +38, South Africa +36, Singapore +24, Japan +22, Mexico +22, Poland +21, Germany +20, Canada +18, the UK +17, France +15, and the US +2.
The deck attributes the US position substantially to partisan divergence. The US has the widest partisan gap of the fifteen: Democrats −10 against Republicans +23, a gap of 32 points. Japan is next at 25 (DPFP +15, LDP +40), then France at 20 (Rassemblement National +16, Ensemble +36) and Mexico at 20. Poland's gap is 1 point. In every other country both major parties are net positive; the US Democratic figure is the only negative party score in the set.
The 2026 trend
On the POLITICO Poll's local question, net support among all adults fell from +9 in January 2026 to −17 in July 2026. The decline is present in every subgroup but not evenly: Trump 2024 voters fell from +19 to 0, Harris 2024 voters from +3 to −30, and 2024 non-voters from +1 to −21.
Beliefs about data centers moved in the same period. The share saying it is true that data centers raise people's electricity bills rose from 43% to 58%; that they deplete the water supply, from 34% to 55%; that they are bad for the planet, from 32% to 49%. The one belief that did not rise is the benefit claim: that building more data centers enables more complicated AI fell slightly, from 62% to 59%.
Party identification followed. Asked which party would want to stop a data center being built locally, the Democrat-minus-Republican margin among all adults widened from +12 to +20 between January and July 2026, driven by Harris 2024 voters, whose margin moved from +27 to +55.
Attitudes to AI itself moved negative over a longer horizon. Net "make better minus make worse" scores across three annual waves run: you personally +10, +15, +5; your career +7, +8, 0; most people you know +7, +12, +2; the economy +5, +6, −4; society +3, +5, −9; and the next generation +6, +10, −4. Every domain except "you personally" is at or below zero in 2026, having been positive in both prior years.
Opposition tracks AI attitudes, not general opposition to development
The deck's most-cited finding is a comparison of opposition to a data center against opposition to the average of the nine other developments, by subgroup. Among all adults, 46% oppose a data center against 23% for the average other development, a gap of 23 points. Among those who believe AI will make society worse, the figures are 73% and 32%, a gap of 41 points. Among those who both believe AI will make society worse and say infrastructure nearby has a negative impact on them, 74% and 36%.
The comparison groups are the point. Respondents who agree "I don't want new development in my area" oppose a data center at 61% and other developments at 35% — a gap of 26 points, smaller than the gap for AI pessimists. Those who say government should respect residents' wishes even if it slows infrastructure are at 57% and 27%; those who agree "I have no control over what gets built in my area" at 54% and 27%. The deck's inference is that opposition is not tracking "straightforward 'NIMBYism' alone." The footnote records that the groups overlap and are not mutually exclusive.
The same pattern appears in net scores. Splitting respondents by whether they say AI will make society worse, the gap for a new data center is 68 points (+5 against −63) — far larger than for a new power plant (−39), an office complex (−38), a warehouse (−33), a shopping mall (−32), a road (−27), a train station (−25), a wind farm (−21), housing (−16) or a solar farm (−15). AI pessimists are more negative about everything; they are much more negative about data centers specifically.
What respondents say would change their minds
Asked about policy announcements from a hypothetical state candidate, the highest-scoring measures are cost-shifting and local control: requiring data centers to pay the full cost of their own electricity and any new power lines (+57), giving local residents a binding vote (+52), a temporary pause on new construction until there are rules on power, water and bills (+48), and minimum distance requirements from homes and schools (+46). Repealing existing tax breaks scores +34 and a permanent state ban on new data centers scores +20. Two pro-build measures score negative: offering property tax discounts (−7) and preventing counties and towns from blocking data centers so the state decides (−9). Fast-tracking planning permission scores 0.
Asked which developer commitments matter most, respondents rank guaranteeing local electricity bills will not rise (+19.3), generating all power on site (+16.0), cooling systems needing almost no water (+12.7) and a yearly household bill credit (+11.7) at the top, and community amenities at the bottom: a new park and sports fields (−18.7), setbacks and screening (−16.2), demolition and land-restoration guarantees (−13.2) and a binding noise limit (−11.4). Among respondents who oppose a data center in their community the resource commitments score higher still — on-site power generation +24.9, bill guarantees +24.5, low-water cooling +20.8 — while the amenity offers score lower, a new park at −25.0.
A conjoint experiment showing two proposals side by side with the option to leave the land undeveloped ranks the attributes by how much they move choice. Water use has the widest span, 8.3 points: a proposal using as much water as 10,000 homes is chosen 26.4% of the time against 34.7% for one using almost no water and returning more than it takes. Permanent jobs span 7.7 points (10 jobs 27.3%, 300 jobs 35.0%); who is building it spans 6.5 points (foreign-investor-owned 28.9%, locally owned in-state 33.4%); what local opposition says spans 4.9 points, with air pollution the most damaging claim at 27.6%; power source spans 3.9 points; company promises span 2.9 points; and construction jobs span only 2.1 points. The average proposal is chosen 30.7% of the time.
Asked openly for the main benefits and drawbacks of building data centers in the US, with up to three selections each, the top drawbacks are higher household electricity bills (32%) and increased risk of water shortages (32%), followed by land use (24%) and noise and light pollution (21%). The top benefits are powering everyday online services (28%), helping American technology companies grow (26%), keeping American data in America (23%) and creating jobs (23%). Staying ahead of China on AI is selected by 16% and national security by 14%; 15% say there are no benefits and 7% say there are no drawbacks.