Seven in ten Americans oppose building AI data centers in their local area, and 48 percent are strongly opposed, according to Gallup. Barely a quarter are in favor. The finding matters commercially this week because reporting attributed to CNBC says Anthropic's confidential IPO filing is expected to name public and political backlash against AI and data centers as a risk factor -- an unusual disclosure for a company whose growth depends on that construction continuing.
Key facts
- 70 percent of Americans oppose local AI data center construction; 48 percent strongly oppose. Only 7 percent strongly favor.
- The finding comes from a March 2-18, 2026 Gallup survey, the first time Gallup has asked the question.
- Half of opponents cite excessive resource use, with 18 percent each naming water and energy.
- Primary source: Gallup.
Two stories usually get told separately: the enormous capital pouring into AI infrastructure, and the local fights over specific buildings. Gallup's numbers are what happens when you put them in the same frame.
The scale of opposition is the part worth sitting with. Seventy percent is not a controversy; it is a consensus, and it is remarkably flat across the usual dividing lines. Gallup reports "no meaningful differences in total opposition by age, race, education, income or urbanicity." Regional variation is mild -- 63 percent in the West and 68 percent in the East against 76 percent in the Midwest and 75 percent in the South. The one substantial gap is environmental: 78 percent opposition among people who worry about environmental quality versus 52 percent among those who do not, which is a much wider split than the one between people worried and not worried about energy costs.
The reasons, gathered through an open-ended follow-up, are concrete rather than ideological. Half of opponents point at resource consumption -- 18 percent name water specifically, 18 percent name energy, 16 percent name pollution including noise. About a fifth cite quality of life: traffic, population growth, land that could be used otherwise. A similar share cite economics running the wrong way -- higher utility bills, cost-of-living pressure, the possibility of taxpayer money underwriting construction. Supporters are simpler to summarize: two-thirds cite economic benefits, and 55 percent name jobs specifically.
The mechanism is not complicated. A large AI data center is a building that draws power at industrial scale and needs water to shed the heat that power becomes. Both come from a shared local supply. Ground Truth has covered the supply side directly -- the United States now tracks more AI data center demand than its average power output -- and the reason that demand keeps growing is that inference is fundamentally memory- and power-bound, so serving more users means more silicon drawing more current, not just better software.
Gallup's own read on the consequences is direct: "Overcoming this opposition stands as a major hurdle in the expansion of AI computing. The intensity of opposition means that proposed data centers are likely to spur grassroots activism from local residents as well as legal challenges." It adds that AI infrastructure "could become an important campaign issue in local and state elections this year, and politicians who favor data centers in their area are likely taking a politically risky stance."
That is the context for the financial story. A syndicated reprint of CNBC reporting says Anthropic confidentially filed to go public in June 2026, and that its prospectus is expected to treat public and political backlash against AI and data centers as a key risk factor. The same report says CFO Krishna Rao has been asked in confidential investor meetings about competition, margin pressure from open-source models, and what happens if data center construction slows.
Why it matters: risk factors in a prospectus are lawyers writing down the things that could break the business, and they are usually a rear-view mirror on what investors already fear. A frontier lab naming public opposition alongside competition and margin compression means it has been asked about it enough times to require an answer. The margin-pressure item is not hypothetical either -- price cuts are arriving from every direction and capable coding models now fit on a 24 GB consumer card.
The honest caveat is significant and worth stating plainly. The Gallup finding is fully verified and directly published. The Anthropic portion is not: the filing is confidential, no public prospectus exists, the reporting rests on unnamed sources, and the version available here is a syndicated reprint rather than the original. Nobody outside the process can check the risk-factor language. Treat the poll as fact and the filing as reporting, and note the gap between the survey window in March and this week's coverage -- the numbers describe the spring, not necessarily today.
Originally published on Ground Truth, where every claim is checked against the primary source.
