The warning sign for health systems in 2026 is not simply that people are nervous about clinical AI. It is that public opposition to AI data centers and healthcare AI adoption are now moving through the same trust environment. Gallup reported in March that 71% of Americans oppose data center construction in their local area, while Reach3 Insights and Rival Technologies found that trust in healthcare AI fell from 52% in 2024 to 44% in 2026.[1][2]
Those two findings do not prove that data center fights caused the healthcare AI trust decline. They do something more operationally useful: they show that hospital AI programs are entering a civic atmosphere in which “AI” no longer sounds like only software, efficiency, or decision support. For many residents, it now also means higher energy demand, water use, pollution concerns, tax incentives, and distant corporate benefit.
| Signal | What it measures | Why it matters for healthcare AI |
|---|---|---|
| 71% oppose local data center construction | Local siting resistance | AI infrastructure is becoming a visible community burden.[1] |
| Healthcare AI trust fell from 52% to 44% | Public confidence in clinical AI | Trust is declining during the same period that infrastructure concern is becoming more prominent.[2] |
| 88% of current healthcare AI users trust it, versus 38% of non-users | Experience gap | Direct exposure may build confidence, but the unexposed public is much harder to reach.[2] |
The Angus Reid Finding Is a Link, Not a Verdict
The most important evidence tying these issues together comes from Angus Reid USA, which places declining trust in healthcare AI alongside public awareness of the resource demands behind AI systems, including water use, energy demand, and reliance on fossil fuels.[3] That matters because it moves the discussion beyond the usual clinical questions of model accuracy, privacy consent, and physician oversight.
The careful reading is that Angus Reid supports a credible spillover mechanism, not a completed causal chain. The study does not establish that a town hearing about a data center directly changes a patient’s willingness to accept an AI-supported diagnosis. But it does show that healthcare AI skepticism is being interpreted in a public conversation where AI’s physical costs are increasingly legible.
That distinction is not academic. If executives treat the decline in trust as only a clinical communications failure, they will respond with patient FAQs, physician champions, validation summaries, and privacy language. Those are still necessary. They are not sufficient if the public is also asking why the technology that promises better care appears to require new industrial-scale infrastructure whose costs are borne locally.
Awareness of Data Centers Is Not Neutral
Pew’s January 2026 findings help explain why the spillover risk is likely to grow as more people learn what sits behind AI services. Among Americans who had heard “a lot” about data centers, 67% said they are mostly bad for home energy costs and 63% said they are mostly bad for the environment.[4] Awareness, in that slice of the public, is associated with more negative judgment.
That is a difficult pattern for health systems because AI adoption strategies often assume familiarity will calm people down. In clinical settings, that can be true when a patient sees a narrow tool used responsibly by a trusted clinician. In infrastructure politics, familiarity may work in the opposite direction: the more residents learn about power demand, water use, and site selection, the more they may see AI as an extraction project rather than a public good.
The opposition is no longer theoretical. Brookings, citing Data Center Watch, reported that in the first three months of 2026, 75 data center projects worth $130 billion were blocked or delayed, equal to the total for all of 2025.[5] That scale gives local disputes a national storyline: communities are not merely asking for better permitting paperwork; they are slowing down major AI infrastructure.
The infrastructure concerns are concrete enough to travel. Harvard Gazette reported that a single hyperscale data center can consume 1.4 GW, comparable to the electricity use of 1 million households.[6] Harvard Chan School researchers also estimated that air pollution from on-site power generation at one Virginia data center could produce $53 million to $99 million in annual health damages, including 3.4 to 6.5 excess premature deaths.[7]
Those figures change the emotional register of AI. The issue is no longer whether a chatbot gives a clumsy answer or whether a radiology model improves workflow. It is whether a community believes it is being asked to absorb energy costs, environmental risk, and governance complexity so that institutions elsewhere can claim innovation.
TIME’s 2026 reporting shows that this opposition has political force. It described bans in Holyoke, Massachusetts; Monterey Park, California; and Seattle, Washington; the ousting of half a city council in Festus, Missouri; and a three-year tax-break moratorium in Arizona.[8] For hospital leaders, the lesson is not that every AI project will be judged like a data center. It is that AI has acquired a local politics that healthcare organizations do not control.
Why This Spills Into the Clinic
Healthcare AI already had its own trust problems before data centers became a household concern. Clinicians and patients worry about hallucination risk, data privacy, and dehumanization; Philips’ 2026 Future Health Index also surfaced concern about whether AI will weaken the human relationship at the center of care.[2][9]
The connection to data center backlash is not that these are the same complaint. It is that they rhyme. In both cases, people are reacting to opacity and asymmetry: someone else builds the system, someone else captures much of the benefit, and the person closest to the consequence is asked to trust that safeguards are adequate.
A nurse explaining an AI-assisted triage tool is not responsible for a hyperscale facility’s power contract. A CMIO validating a sepsis model is not responsible for a city council’s fight over water use. But patients and clinicians do not necessarily sort those issues into separate governance boxes. The brand in front of them is “AI,” and the trust account is shared more than many vendors would like to admit.
The Counter-Signal: Use Can Build Trust
The strongest reason not to overread the backlash is the same Reach3/Rival finding that exposes the trust problem: 88% of current healthcare AI users trust the technology, compared with 38% of non-users.[2] That is a very large gap, and it suggests that direct experience with a bounded, useful healthcare AI tool can build confidence.
For health systems, this is the practical opening. Trust may be recoverable when AI is visible as a specific clinical aid, used by accountable professionals, with a clear path for review and correction. A scheduling assistant, documentation support tool, or diagnostic aid does not have to carry the full symbolic burden of the AI economy if the organization can show who uses it, what it is allowed to do, what it is not allowed to do, and who answers when it fails.
But that opening is narrowest among non-users, and non-users are exactly the group most likely to encounter AI first through headlines about energy demand, local opposition, and opaque infrastructure deals rather than through a helpful clinical experience. The health system that waits until rollout day to explain its AI program is entering the conversation late.
What Health System Leaders Should Take From This
The boardroom implication is straightforward: healthcare AI trust can no longer be managed only inside the clinical evidence file. Validation studies, physician education, privacy notices, and patient-facing explanations remain central. They answer whether a tool is safe, useful, and governed in care delivery. They do not answer whether the public now associates AI with a broader set of civic costs.
- AI governance committees should track public infrastructure concerns as part of adoption risk, not only model performance and regulatory exposure.
- Patient communication should name the boundaries of each tool rather than asking people to trust “AI” as a general category.
- Vendor diligence should include questions about compute sourcing, environmental disclosures, and infrastructure dependencies when those issues could affect public confidence.
- Clinician champions should not be left to defend an entire technology stack; they need institution-level answers about accountability and external impacts.
Data center backlash is best understood as an amplifier of healthcare AI skepticism, not the sole cause of it. The decline in trust also reflects familiar anxieties about error, privacy, dehumanization, and corporate control. What has changed in Q3 2026 is that AI’s physical footprint has become harder to separate from AI’s clinical promise. A credible healthcare AI strategy now has to account for both.
References
- Americans Oppose Data Centers in Their Area, Gallup, March 2026
- The Trust Gap Between AI Users and Non-Users in Healthcare, Healthcare IT Today, July 13, 2026
- AI in Healthcare: Trust Decline, Angus Reid USA, 2026
- How Americans View Data Centers’ Impact in Key Areas, From the Environment to Jobs, Pew Research Center, March 12, 2026
- Data Center Backlash Signals a Fight Over AI Power, Brookings, July 2026
- Why Are Communities Pushing Back Against Data Centers?, Harvard Gazette, April 2026
- Analyzing Air Pollution, Health, Economic Risks From AI Data Centers, Harvard T.H. Chan School of Public Health
- AI Elections Data Center Backlash, TIME, June 18, 2026
- AI Healthcare Trust Gap 2026 Future Health Index Philips, Medical Daily
Comments
Join the discussion with an anonymous comment.