From the Interstate Highway Era to AI Data Centers: Safeguarding Housing and Health in Infrastructure Placement
Michael Akinwumi
Chief AI Officer & Head of the Responsible AI Lab
National Fair Housing Alliance (NFHA)
Today, the U.S. stands at another infrastructure inflection point: not asphalt and concrete, but fiber, substations, and data centers. The AI Action Plan positions digital infrastructure as a pillar of national competitiveness and geopolitical leadership.[1] Just as the Interstate Highway System, a significant infrastructure project established via the Federal-Aid Highway Act of 1956, was justified based on the need to modernize our travelways, enhance national defense, and promote economic growth. Highways promised mobility, suburban expansion, and commercial development. Yet routing decisions often cut directly through Black neighborhoods, displacing families, depressing property values, disconnecting people from job and economic centers, and fragmenting social networks. These projects were frequently framed as neutral exercises in engineering and progress. The harms were not always explicitly articulated as intentional discrimination. But their effects were durable: concentrated wealth extraction, reduced access to opportunity, and generational racial disparities in housing and health.[2]
Federal momentum toward AI dominance coincides with regulatory shifts that attempt to concentrate federal AI policy, weaken environmental guardrails, and usurp local- and state-level authority to make AI infrastructure decisions. Hence, the question is not whether AI infrastructure should be built. It is how and where it is built, lest we repeat the intentional displacement and economic harm inflicted on Black communities in the 1950s. AI infrastructure projects should proceed only after a mandatory environmental and social impact assessment and the adoption of a legally binding community benefit agreement, ensuring that marginalized communities are not disproportionately burdened by their development.
AI data centers are expanding rapidly across the country to meet escalating computational demand, and their environmental impact is clearer every day. AI systems could generate between 32.6 and 79.7 million tons of CO₂ emissions and consume 312.5–764.6 billion liters of water in 2025 — levels comparable to New York City’s annual carbon output and roughly equivalent to the world’s yearly bottled‑water consumption.[3] Water consumption at hyperscale facilities can reach millions of gallons per day for cooling alone, exacerbating local water scarcity.[4],[5] These facilities require substantial site footprints that can contribute to resource-intensive infrastructure expansion.[6] Depending on their design and construction, data centers can also release toxins into nearby communities with devastating health, economic, and other impacts.[7]
Decisions on where large-scale AI data centers are ultimately located carry material consequences for property values, health outcomes, housing affordability and accessibility, and public services. Industrial adjacency and environmental stigma can influence real estate markets. Research shows that proximity to pollution sources is not randomly distributed: Black residents face 1.54 times higher particulate matter (PM2.5)[8] burdens than the overall population, with those in poverty facing 1.35 times higher impacts.[9] These disparities operate across national, state, and county levels. Environmental risk can depress home values, and because housing is the primary asset for most American families, property value suppression translates directly into intergenerational wealth loss. Health burdens compound this dynamic. Exposure to particulate matter is associated with cardiovascular and respiratory disease.[10] While all of the causes of asthma remain unclear and may be related to increased exposure to pollution, in 2019, non-Hispanic Black individuals were almost three times more likely to die from asthma-related causes than their non-Hispanic White counterparts, and non-Hispanic Black children faced an even more severe disparity, with asthma death rates eight times higher than those of non-Hispanic White children.[11]
Civil rights law provides a framework for understanding these structural dynamics. The Fair Housing Act’s disparate impact doctrine recognizes that policies may violate the law when they produce unjustified discriminatory effects, regardless of intent. Environmental justice scholarship has long demonstrated that hazardous facilities disproportionately burden people of color and low-income neighborhoods. Studies of hazardous waste siting and air pollution reveal
systematic targeting or tolerance of risk in communities with less political power.[12] Cheryl
Harris’s theory of “whiteness as property” underscores how property rights and racial hierarchy have historically intersected to secure advantage for some communities while constraining others.[13] Decisions on where AI data centers are located are no different. Even facially neutral criteria can replicate patterns of structural vulnerability.
The risk of repeating harms to marginalized communities based on infrastructure decisions is clear. Highway construction was justified as a national imperative. Newly-constructed highways and bridges often ran through communities with the least political leverage — not by accident, but by design. Local governments and industry allow data centers to be placed in locations where land is cheaper and historically divested communities have less concerted political power to push back on the decisions. Generous tax incentives may attract facilities without guaranteeing durable local benefit. Environmental and grid costs can be externalized onto surrounding communities. The through-line may not be explicit racial intent; it is the predictable distribution of burdens along lines shaped by past racial segregation and ongoing disinvestment in marginalized communities.
This moment calls not for opposition to AI infrastructure, but for integrating fair housing analysis, environmental burden mapping, and housing impact modeling into infrastructure permitting decisions and design. Infrastructure is never neutral; it can either build or break communities. Civil-rights–informed decisions on how the burdens of innovation will be shouldered for large-scale AI infrastructure should include:
- Mandatory cumulative impact assessments. Evaluating how the design, construction, and resource needs of new data centers interact with and impact existing environmental burdens in a neighborhood.
- Fair housing impact assessments. Modeling how large-scale infrastructure projects will affect local land prices, utility rates, housing accessibility, and housing affordability before permits are granted.
- Community benefit agreements. Tying tax abatements and land-use approvals to enforceable commitments for local infrastructure improvements and funding to promote fair and affordable housing.
The Interstate Highway Era taught us that when “national interest” is defined without a civil rights lens, communities that are already marginalized pay the highest price. As we build the infrastructure of the AI age, we have a responsibility to ensure that the “golden age of human flourishing” promised by technological dominance is not built upon the displacement, isolation, economic destruction, and disinvestment of the same communities we have failed before. We must apply these lessons now, before new choices become embedded for generations.
[1] Michael J. Kratsios, David O. Sacks, and Marco A. Rubio, “Winning the Race: America’s AI Action Plan,” The White House (July 2025), https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf.
[2] Deborah N. Archer, “White Men’s Roads through Black Men’s Homes: Advancing Racial Equity through Highway Reconstruction,” The Vanderbilt Law Review (2020), https://scholarship.law.vanderbilt.edu/vlr/vol73/iss5/1.
[3] Alex de Vries-Gao, “The Carbon and Water Footprints of Data Centers and What This Could Mean for Artificial Intelligence,” Patterns (2026), https://doi.org/10.1016/j.patter.2025.101430.
[4] Yu Tao and Peng Gao, “Global Data Center Expansion and Human Health: A Call for Empirical Research,” Eco-Environment & Health (Sept. 2025), https://doi.org/10.1016/j.eehl.2025.100157.
[5] David Mytton, “Data Centre Water Consumption,” npj Clean Water (Feb. 15, 2021), https://doi.org/10.1038/s41545-021-00101-w.
[6] Can Hankendi, Ayse K. Coskun, and Benjamin K. Sovacool, “Why Transparency Matters for Sustainable Data Centers and Carbon-Neutral Artificial Intelligence (AI),” iScience (Nov. 21, 2025), https://doi.org/10.1016/j.isci.2025.113705.
[7] The Kapor Foundation, “The Unequal Burden of Data Centers: An Examination of the Environmental and Public Health Impacts on Communities in California,” (Dec. 9, 2025), https://kaporfoundation.org/datacenters-envt-health.
[8] Particulate matter (PM) refers to a mixture of solid particles and liquid droplets suspended in the air that are small enough to be inhaled and, in some cases, absorbed into the bloodstream. PM2.5 refers to particulate matter 2.5 micrometers or smaller, linked to respiratory and cardiovascular disease and increased risk of premature death.
[9] Ihab Mikati, Adam F. Benson, Thomas J. Luben, Jason D. Sacks, and Jennifer Richmond-Bryant, “Disparities in Distribution of Particulate Matter Emission Sources by Race and Poverty Status,” The American Journal of Public Health (March 7, 2018), https://doi.org/10.2105/AJPH.2017.304297.
[10] Id.
[11] Office of Minority Health, “Asthma and African Americans,” U.S. Department of Health and Human Services (Feb. 2021), https://minorityhealth.hhs.gov/node/39/revisions/39/view.
[12] Paul Mohai and Robin Saha, “Which Came First, People or Pollution? Assessing the Disparate Siting and Post-Siting Demographic Change Hypotheses of Environmental Injustice,” Environmental Research Letters (Nov. 18, 2015), https://doi.org/10.1088/1748-9326/10/11/115008.
[13] Cheryl I. Harris, “Whiteness as Property,” The Harvard Law Review (June 1993), https://www.jstor.org/stable/1341787.
Dr. Michael Akinwumi is an executive leader at the intersection of technology, civil rights, and equity — working to ensure that the systems shaping tomorrow do not replicate the injustices of yesterday. Grounded in a computational justice framework, he guides organizations in developing and deploying technology toward solutions that meet rigorous standards of alignment, safety, security, and policy compliance. Michael is widely recognized for strengthening institutional standards for trustworthy technology deployment, guiding organizations integrating alignment, risk management, and accountability into operational AI systems. He partners with policymakers, industry leaders, and community stakeholders to advance innovation that protects and expands opportunity for communities that have historically been marginalized.