ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
TopicDisconfirmAI EditorialUpdated 2026-11-22 12:55

AI Data Center Racial Disparity

2 steps1 pieces2 Sources

A new study by researchers at George Washington University and American University reveals that data centers are disproportionately located in communities of color across the United States. Analyzing 81,000 census tracts and 5,291 data center locations, the report "Redlining in the Age of AI" found that majority-Black neighborhoods have 2.2 times the odds of hosting a data center compared to white neighborhoods in the same state. The disparity is even greater for Hispanic (2.6x) and Asian (4.2x) neighborhoods. These findings persist even after controlling for economic factors like income, education, land costs, and proximity to transmission lines.

The researchers attribute this pattern partly to "infrastructural inheritance," noting that high-voltage transmission lines were historically built through Black neighborhoods due to redlining practices. While developers may not intentionally target these areas, race remains deeply embedded in location decisions. The study also highlights a divergence: large hyperscaler campuses (like those from Meta or Google), which often promise significant local economic benefits such as jobs and tax revenue, tend to be built in rural white communities. In contrast, smaller mixed-use facilities and network hubs are more likely to appear in urban communities of color, where residents may face environmental burdens without receiving comparable economic returns.

To address this inequity, the authors recommend implementing data center-specific zoning regulations, prioritizing racial equity in planning discussions, and requiring racial impact statements before construction begins. They argue that tech policy must involve negotiation between companies and affected communities, recognizing that AI has a physical geography shaped by historical social and political forces.

How it got here

  1. 2026-11-19

    Nov 19, 2026: Researchers from George Washington University and American University publish the study 'Redlining in the Age of AI,' finding that majority-Black neighborhoods have 2.2 times the odds of hosting a data center compared to white neighborhoods in the same state.

    404media.co ↗
Sign in / Sign up
ReadingResearchRadarInvestment framework
  • 2026-11-19

    Nov 19, 2026: The study reports that disparities increase to 2.6 times for majority-Hispanic neighborhoods and 4.2 times for majority-Asian neighborhoods, persisting even after controlling for population, income, education, land costs, and proximity to transmission lines.

    aiweekly.co ↗
  • What is still unsettled

    • Will federal or state regulators adopt the study's recommendation for mandatory racial impact statements for new data center projects?
    • How will the distinction between hyperscaler campuses and smaller co-location facilities influence future zoning laws and community benefit agreements?

    Sources

    • 404media.co404 Media↗
    • aiweekly.coAI Weekly↗

    Coverage in this topic

    • US Study: Black Neighborhoods Twice as Likely to Host Data Centers2026-11-19News

      Researchers from George Washington University and American University find that minority neighborhoods are disproportionately selected for AI data center construction, even after controlling for economic factors.