Black neighborhoods have 2.2 times the odds of hosting data centers, study finds
An analysis of 81,000 U.S. census tracts finds disparities that persist after accounting for income, land costs and power infrastructure. The authors recommend racial impact reviews before construction.
A nationwide analysis of 81,000 census tracts and 5,291 data centers found racial disparities in where facilities are located that persisted after researchers accounted for income, land costs, power access and other factors. The authors say historical infrastructure patterns, including transmission lines built in majority-Black neighborhoods, may help explain the results; they do not accuse developers of deliberately targeting communities of color. Their recommendations focus on addressing racial equity before projects are sited or built.
01
Within the same state, majority-Black neighborhoods had 2.2 times the odds of hosting a data center as white neighborhoods; the ratios were 2.6 for Hispanic and 4.2 for Asian neighborhoods.
02
The analysis includes hyperscalers as well as smaller co-location facilities, so its findings extend beyond large AI-focused data center projects.
03
The researchers recommend data-center-specific zoning, racial equity considerations in siting decisions and required racial impact statements before construction.
The geography of computing infrastructure follows racial divides that income and land prices do not fully explain. A new study from George Washington University and American University researchers finds higher odds of hosting data centers in Black, Hispanic and Asian neighborhoods—even after accounting for economic conditions and access to power infrastructure.
The study, Redlining in the Age of AI: The Relationship between Data Center Locations and Race in the US, examined 81,000 U.S. census tracts and location data for 5,291 data centers. Matthew Gault reported the findings and interviewed coauthors Samantha Bradshaw and Dhanaraj Thakur for 404 Media on November 19.
Separating race from other siting factors
The researchers did more than compare neighborhood demographics with a map of facilities. Their analysis accounted for population, household income, education, land costs and proximity to major transmission lines. It also accounted for differences between states. The racial disparities remained after those adjustments, the researchers wrote.
They also found that the pattern held despite data centers’ concentration in and around metropolitan areas, which are often more racially diverse. Across the analysis, the odds of hosting a facility rose as the share of Black, Hispanic and Asian residents increased. Majority-Black neighborhoods had 2.2 times the odds of white neighborhoods in the same state.
Odds of hosting a data center
2.2×Majority-Black neighborhoods
The study found 2.2 times the odds of hosting a data center compared with white neighborhoods in the same state.
2.6×Majority-Hispanic neighborhoods
The study found 2.6 times the odds of hosting a data center compared with white neighborhoods.
4.2×Majority-Asian neighborhoods
The study found 4.2 times the odds of hosting a data center compared with white neighborhoods.
Power access carries a history
Bradshaw and Thakur told 404 Media that stories about data-center pollution affecting Black and Brown communities in Memphis prompted their investigation. They wanted to examine whether those local concerns fit a nationwide pattern, rather than treating individual projects as isolated cases.
The researchers point to the legacy of redlining—the denial of loans and other opportunities that helped confine Americans of color to particular neighborhoods—as part of the explanation. Access to electricity offers one concrete connection: developers often seek sites near transmission lines, and those lines were often built in majority-Black neighborhoods.
That explanation is not an accusation that developers deliberately target communities of color. Thakur said he doubted that was their intention. His argument instead concerns how race is embedded in the other factors shaping site selection. The study’s statistical finding and the researchers’ explanation of its historical roots are related, but distinct.
Different facilities, different community bargains
The study also distinguishes among types of data centers. Its scope includes both hyperscalers—the enormous facilities associated with companies such as Meta and Google—and smaller mixed-use buildings. The latter can combine computing capacity with network transmission hubs, which the study describes as co-location facilities. The findings therefore concern more than the largest AI megaprojects.
That distinction complicates the picture of who hosts infrastructure and who is promised its benefits. According to the study as described by 404 Media, most hyperscale projects promising major economic benefits are built in white rural communities. Those promises should not be confused with measured benefits across all the neighborhoods in the analysis.
The authors recommend bringing racial equity into siting decisions before construction begins. Their proposed measures include:
Zoning rules written specifically for data centers.
Explicit consideration of racial equity in discussions about where facilities are built.
Required racial impact statements before projects are constructed.
Bradshaw and Thakur frame technology policy as a negotiation between companies and the communities that use their technology. Their proposed response puts that negotiation at the siting stage, when decisions about a facility’s location are still being made—not only after it becomes a neighborhood presence.
Sources
404media.coBlack Neighborhoods Are Twice as Likely To House a Data Center, Study Finds
Reader comments
Newest comments first. Replies stay oldest first.