Research institute · Pew Research Center
67% of planned US data centres are in rural areas
Pew Research Center's analysis of Data Center Map records finds the US data centre footprint is shifting decisively rural. While 87% of existing centres sit in urban counties, 67% of the 1,500-plus planned builds are heading to rural ones, a structural shift in where AI infrastructure lives.
Source published Summarised
Key findings at a glance
- 67% of planned US data centres will be built in rural counties, inverting the existing footprint, where 87% sit in urban areas.
- 1,500+ new data centres are in development across the US, on top of the 3,000-plus already operational. The build-out is a step-change, not a marginal expansion.
- 39% of planned data centres are going into counties that currently host none, opening hundreds of new local economies to the AI infrastructure build.
- 48% of all planned US data centres will be built in the South alone. The South and Midwest together account for three-quarters of new construction.
- 62% growth in data centres expected in the South region, 754 planned added to 1,209 existing facilities.
- 64% growth in data centres expected in the Midwest, 419 planned on top of 655 existing. The fastest regional growth rate in the country.
- 287 planned data centres in Virginia, the highest of any state. Texas follows at 170, Georgia at 141, Illinois at 123.
- 38% of Americans currently live within five miles of an operational data centre. The structures are not abstract; most of the country already lives near one.
- 90% of US data centres are within five miles of at least one other. They cluster heavily rather than spreading evenly across the map.
- 42% of the US population will live within five miles of an existing or planned data centre once the pipeline completes.
- 8,512 US adults surveyed by Pew between 20 and 26 January 2026 for the parallel public-opinion analysis on data centre awareness.
What the report covers
This Pew Research Center analysis looks at where the next wave of US data centres is being built, using Data Center Map records current to 19 February 2026. Pew combined facilities listed as 'under construction', 'planned' and 'land banked' into a single planned category and stripped out anything marked 'cancelled' or 'uncertain'. County-level cuts use the US Census Bureau's urban–rural classification.
The headline is that the rural–urban split inverts from the existing footprint. Of the roughly 3,000 operational US data centres, 87% are in urban counties. Of the 1,556 facilities in the planning, construction or land-banked pipeline, 67% are in rural counties. The South leads on absolute volume, 754 planned facilities, with Virginia (287) and Texas (170) the two largest state pipelines. The Midwest is the fastest-growing region in percentage terms, with planned builds equivalent to 64% of its existing base.
Pew pairs this geographic analysis with a survey of 8,512 US adults examining awareness and opinions about data centres. The headline survey finding is that proximity does not materially shift attitudes, Americans who live near a data centre hold similar views to those who do not. That detail matters because the AI capex narrative tends to assume location-led backlash; the data so far does not support that.
What this means for UK marketers
For UK B2B technology vendors, agencies and content teams selling into the US AI infrastructure build-out, this is the geography layer behind the capex headlines. Three implications worth pricing into 2026 plans.
The buying centre is splitting. The named accounts on a US enterprise data centre programme still sit in cities, corporate procurement, finance, the CTO office. But the operational stakeholders, site selection, energy procurement, water and cooling, physical security, local-authority liaison, are increasingly in rural counties from Virginia, Texas, Georgia and Ohio outwards. UK vendors that have historically marketed only to urban-HQ named accounts should audit their buying-committee map. The B2B campaigns that work for this audience look more like trade-press placements, county-level partnerships and field-marketing events than the LinkedIn-only model that suits city-HQ targets.
The 'AI infrastructure' story needs new imagery. Most UK creative and PR currently illustrates the AI capex story with glass-and-steel urban skylines or stock-photo server racks. The actual physical reality is large windowless warehouses on the edge of rural counties. Brands publishing thought leadership on AI energy, AI water use, AI labour markets or AI community impact need imagery and on-the-ground reporting that matches where this build-out is actually happening. The Pew finding that 39% of planned centres go into counties with none today is the most useful single statistic for grounding that pivot.
The UK has a parallel set of decisions in flight. The British government's AI Opportunities Action Plan sets out a UK data centre and compute strategy, including the AI Growth Zone concept and the designation of compute as critical national infrastructure. UK consultancies, planning advisors and energy-sector marketers will see analogous rural-versus-urban location debates play out at smaller scale here over the next 24 months. The US data is a working preview of which stakeholders end up shaping the conversation and which messaging frames travel.
One reassuring data point for anyone selling against community-backlash anxiety: Pew finds that Americans who already live near a data centre hold roughly the same opinions on their environmental, jobs and home-energy impact as those who do not. Familiarity is not currently shifting sentiment in either direction. That is a different, and more workable, earned-search and reputation position than the framing most agencies have been operating against.
How they got the data
Data on the number and location of US data centres comes from Data Center Map, one of the longest-running industry databases in the sector, current as of 19 February 2026. Pew combined facilities listed as 'under construction', 'planned' and 'land banked' into a single planned category and removed projects marked 'cancelled' or 'uncertain'. Geographic analysis uses the US Census Bureau's urban–rural classification and county geographies.
The accompanying public-opinion analysis is a survey of 8,512 US adults conducted between 20 and 26 January 2026, drawn from Pew's American Trends Panel. The ATP uses address-based sampling for recruitment via the US Postal Service's Computerized Delivery Sequence File, with online and live-telephone interviewing in English and Spanish. Oversamples are weighted back to population proportions to produce nationally representative estimates.
Research institute
Read the full report
From Pew Research Center · published 13/04/2026