1,440 data center facilities across the United States as of 2026. Who is building them, where they are going, and what communities are doing.
529 data centers are already operating across the country, powering the infrastructure behind cloud computing, artificial intelligence, and everyday digital services.
119 more are under construction right now. The build-out is accelerating.
649 more have been proposed. If they are built as planned, the number of facilities in the US would nearly double.
Virginia leads all other states in data centers by a wide margin in this dataset. Much of what appears on the map is still only proposed, meaning the footprint five years from now could look very different than it does today.
Click any state bar to view that state. Hover over dots for details.
The six largest operators ranked by total facility count, including everything proposed. For most of them, the majority of what they have staked out has not been built yet.
Amazon appears under two names in the data, “Amazon” and “Amazon Data Services Inc.” Combined, they account for 75 facilities, with 51 located in Virginia. Of these, 42 are proposed, meaning Amazon has nearly twice as much planned capacity as it currently operates. An $11 billion campus in Georgia is under construction, and a proposed $20 billion, 960 MW campus in Pennsylvania is planned next to the Susquehanna Nuclear Plant.
Microsoft has one of the largest operator footprints in the dataset, with 48 facilities listed under its name and 6 more under “Microsoft Corporation.” However, only 13 are currently operating, while 33 are still proposed. These reflect long-term land claims and power agreements made years ahead of construction, with Virginia, Georgia, and Texas as the primary targets.
Google has 10 operating, 5 under construction, and 16 more proposed. A 3:1 pipeline ratio. The $2 billion Project Zodiac in Indiana is under construction. A $3 billion campus in Milan County, Texas is proposed. Google also appears as a tenant in 8 additional facilities operated by others, making its real footprint larger than the operator count shows.
Meta currently has 8 facilities simultaneously under construction, more than any other company in the dataset. The $15 billion Hobart, Indiana campus at 2,400 MW is the largest active build in the data. The Louisiana facility, powered by 3 dedicated natural gas plants at up to 2,200 MW, is reported to be one of the largest data center projects underway. Only 2 more are in the proposed stage.
STACK Infrastructure operates 12 facilities, all in Virginia, all in Loudoun County. One of the more geographically concentrated operators in the dataset. Unlike the hyperscalers expanding nationally, STACK has made a single massive bet on Northern Virginia's power infrastructure and existing fiber density. 4 more are proposed in the same corridor.
Digital Realty is one of the oldest colocation operators in the dataset. 12 of its 16 facilities are already operating. Unlike the hyperscalers building for their own workloads, Digital Realty leases space to tenants including cloud providers. It concentrates heavily in Virginia and Georgia, where it appears in some records as a tenant of its own facilities.
But location is only part of the story. The next question is what kinds of communities these facilities are actually landing in. At the national level, counties with data centers tend to have higher incomes than the average. But zooming into individual states reveals a more specific pattern: facilities are not concentrated in the wealthiest counties. Instead, they cluster in more affordable counties within already wealthy regions, where infrastructure is available but land is cheaper.
Each dot below is a county with at least one data center, plotted by its median household income. The dashed line marks the US county median of $66,062. Most dots fall to the right of it. On its own, this suggests the industry tends to land in well-off places.
But the national view does not show the whole picture. Select a state below and compare the two maps side by side. The counties with the most data centers are almost never the highest-income counties in that state. The industry lands in wealthy regions, but it picks the cheaper ground within them.
Select a state and use the metric buttons to switch views. The left map shows data center density relative to that state. The right map shows median household income on a national scale. Hover any county to see figures for both maps at once.
The scatter plot shows that data center counties tend to sit above the national income median of $66,062. But the pattern becomes clearer when comparing counties within individual states across both panels. The densest data center locations are not necessarily the richest counties. Instead, they tend to be places where infrastructure capacity, power access, and available land overlap within already economically strong regions.
Looking across states, a consistent pattern emerges: data centers are not concentrated in the wealthiest counties. Instead, they cluster in more affordable counties within high-income regions. The numbers reflect this shift. Counties where facilities already operate average $79,771 in median income, while counties with only proposed facilities average $76,341. That $3,430 gap widens further in the pipeline. In the 42 counties where proposed facilities outnumber operating ones by three to one, median income drops to $64,691, below the national average.
171 facilities in the FracTracker database have some form of documented community opposition. 77 have active petitions filed against them and 78 have named advocacy groups on record. Those numbers are almost certainly undercounts.
Looking across the cases in FracTracker, the sequence tends to follow a similar path. A community learns about a proposed facility through a news report or a planning board notice, often well after permits have been filed. Residents organize, start a petition, attend public meetings, and ask questions about water use and power capacity. In most documented cases, the project moves forward regardless. The 77 active petitions and 78 named advocacy groups in this dataset represent real, organized resistance. Most of it is happening without lawyers, without access to environmental assessments, and without awareness of what other communities went through in the same situation.
What connects these cases is not which state they are in or which company is involved. It is the timing. By the time a community is organized enough to mount real opposition, the land has often already been rezoned, the power agreements signed, and the permitting process well underway. The public meeting comes last.
The nine cases below are the most thoroughly documented in the FracTracker database. Each card tells what happened. Hover to read. Eight of them follow the same arc. ONE DOES NOT.
One case in the grid below ended differently. In Adairsville, Georgia, a sustained community campaign against Project Springbank ended with the developer pulling the application entirely. No permit was issued. No facility was built. It is one of the very few cases in the full FracTracker database where organized opposition led directly to a project being stopped. It sits at the center of the grid below not as an exception to dismiss, but as a proof of concept.
Every number in this project was calculated from publicly available records. The full picture is considerably larger, and most of it is not accessible to the public.
848 of 1,440 facilities, or 59%, have no size data in the FracTracker database. Companies are not required to disclose this publicly.
After working through these gaps, one thing became clear. More detailed versions of this information do exist in commercial datasets. Facility sizes, power consumption figures, water use agreements, and cooling system details are often recorded in full elsewhere, but they are not freely accessible. Instead, they sit behind commercial data platforms that require expensive licenses, sometimes costing thousands of dollars per year. These datasets are commonly used by investors and industry analysts, while regulators, journalists, and nearby communities often rely on incomplete public records.
The question the gaps raise is not complicated. If 98% of water use data and 95% of power source data are missing from the public record, and the detailed versions of that information are available behind commercial licenses, who exactly is this information being withheld from? The industry knows. The data brokers know. The communities most directly affected often do not.
Facility data: FracTracker Alliance Data Centers Database (2026). 1,440 facilities with operator, status, location, cost, community pushback, and source URL fields. Available at fractracker.org/data/data-centers/
Income data: Median household income by county, 2019–2023 ACS 5-year estimates via HDPulse (National Institute on Minority Health and Health Disparities, hdpulse.nimhd.nih.gov).
County boundaries: US Census Bureau county shapefiles via us-atlas@3 (topojson/us-atlas on npm). FIPS codes used for income matching.
This project is for informational purposes only and is intended to provide insight rather than make definitive claims. All figures are drawn directly from the sources listed above. Data reflects conditions as of early 2026. Interpretations are based on observed patterns in the data and should not be treated as conclusive findings.