Research institute · Pew Research Center
54% of US Adults Say Data Centres Harm the Environment
Public opinion on data centres has turned sharply negative in seven months: 54% of US adults now say they are mostly bad for the environment, up from 39% in January. Six in ten would be uncomfortable with one operating in their area.
Source published Summarised
Key findings at a glance
- 54% US adults who say data centres are mostly bad for the environment
- 39% Said the same in January 2026, so the negative share rose 15 points in seven months
- 50% Say data centres are mostly bad for home energy costs, up from 38% in January
- 49% Say they are mostly bad for the quality of life of people living nearby
- 4% Say data centres are mostly good for the environment
- 60% Would be not too or not at all comfortable with a new data centre in their area
- 7% Would be extremely or very comfortable with a new local data centre
- 24% Say data centres are mostly bad for local jobs, up from 15% in January
- 21% Say they are mostly bad for local tax revenue, up from 12% in January
- 64% Of 18 to 29 year olds say data centres are mostly bad for the environment
- 42% Of over-65s say the same, the widest age gap in the survey
- 10,548 US adults surveyed on the American Trends Panel, 20 July to 9 August 2026
What the report covers
Pew Research Center published this analysis on 22 September 2026, drawing on a survey of 10,548 US adults fielded from 20 July to 9 August 2026. It is a second wave rather than a one-off reading: Pew asked the same questions in January 2026, so every headline figure arrives with a directly comparable baseline seven months earlier.
Respondents were asked whether data centres are mostly good or mostly bad for five things: the environment, home energy costs, the quality of life of people living nearby, local jobs and local tax revenue. Two further questions covered how comfortable people would be with a new data centre operating in their area, and how much they had heard or read about data centres at all.
All five impact measures moved in the same direction, and none of them moved gently. The sharpest shift was on quality of life nearby, where the share saying "mostly bad" rose 19 points from 30% to 49%. Awareness climbed over the same period: the share who had heard nothing at all about data centres fell from 25% to 12%. That combination rules out the simplest explanation. People are not turning negative because they are uninformed. They are turning negative as the subject becomes more familiar. For the UK reading on attitudes to the technology these facilities exist to run, see our earlier report on how just 36% of UK adults think AI will benefit them.
What this means for UK marketers
The first caveat is the obvious one: this is US data, and UK attitudes are measured separately. It still matters here, because the UK is running the same experiment. The government's AI Growth Zones programme is putting large compute facilities into named communities, so the local planning arguments Pew is measuring are arguments UK regional media will be having as well.
The practical point for anyone writing marketing copy is that the infrastructure behind AI has stopped being neutral background. If your positioning leans on "AI-powered" as a self-evident good, you are now speaking to an audience in which half the people asked believe the physical layer underneath it pushes up their energy bill. Note what the objection is not: almost nobody here is complaining about model quality or about robots taking over. The complaints are electricity, local amenity and cost. Those are kitchen-table objections, and they are much harder to answer with a product benefit than an abstract worry is.
Two things follow. First, be careful with environmental claims. If you market AI or cloud services and you also want credit for being efficient or low-carbon, the CMA's Green Claims Code sets out what you need to be able to substantiate. Vague sustainability language attached to a compute-heavy product is exactly the sort of claim that attracts scrutiny, and the public mood in this survey is the reason scrutiny is increasing. Second, if you have a genuine efficiency story, the audience for it is arriving. Lighter pages, less client-side processing and fewer wasted server round trips are measurable, defensible and cheap to evidence, which is the basis of our low-carbon web design work.
One segmentation note before you apply any of this. The negativity is strongest among 18 to 29 year olds, at 64% on the environment question against 42% of over-65s, and it is strongest among graduates. If your buyers skew young and educated, assume the scepticism sits above the national average rather than at it. Our earlier report on US under-30s being more wary of AI than excited found the same age gradient on a different question, which suggests this is a stable feature of the audience and not a quirk of one survey.
How they got the data
Pew surveyed 10,548 US adults between 20 July and 9 August 2026. Respondents were members of the American Trends Panel, Pew's probability-based online panel recruited through national random sampling of residential addresses, with results weighted so the sample represents the full US adult population. The January 2026 comparison figures come from an earlier wave of the same panel, which is why the two readings can be set side by side. Pew publishes the questionnaire, the detailed responses and the full methodology statement alongside the analysis.
One reading note changes how these percentages should be quoted. Each impact question keeps every respondent in the base, including the 12% who say they have not heard of data centres and the roughly one in five who answer "not sure". The 54% who call data centres mostly bad for the environment is therefore a share of all US adults, not a share of people holding a view. Among those who say they have heard a lot about data centres, Pew's own crossbreak puts that figure at 73%. Quote the 54% as a population figure and the 73% as an informed-opinion figure, and do not mix the two.
Research institute
Read the full report
From Pew Research Center · published 22/09/2026