As billions pour into data center infrastructure across the United States, a familiar pattern is emerging and Black communities are once again at the center of it, without equity in the outcome.
On the eastern edge of a majority-Black neighborhood in the Georgia Piedmont, a construction crane is lifting steel into the autumn sky. The site, twenty acres of formerly flat commercial land purchased eighteen months ago by a Delaware LLC whose beneficial owners took months of public records requests to identify, will become, by 2027, a hyperscale data center campus drawing hundreds of megawatts of electricity from a regional grid that already strains under industrial load. The neighborhood’s residents learned about the project at a county commission meeting, three weeks after the tax abatement was approved. Nobody knocked on their door.
That scene, with variations in geography, corporate identity, and timeline, is playing out across the United States in 2026. The AI infrastructure build-out is the largest wave of industrial investment in American communities since the Interstate Highway System. And the communities most directly affected by it, those bearing the land, labor, energy, and environmental costs of the facilities that power the AI economy, are disproportionately Black and low-income. This is not a coincidence. It is a pattern, and it has a name: sacrifice zone economics.
The numbers are not ambiguous. Research overlaying EPA EJScreen 2024 demographic and pollution indicators against Data Center Map site inventories across five major hyperscaler growth corridors shows that host census tracts score consistently higher on composite environmental burden indicators than their surrounding regions. In the corridors studied, Northern Virginia, the Georgia Piedmont, central Texas, the Phoenix metropolitan area, and the Ohio Valley, majority-minority communities and lower-income communities are overrepresented as data center host locations relative to their share of regional geography. The primary drivers of that pattern are the same drivers that have always shaped American industrial location decisions: cheap land and cooperative local governments hungry for corporate investment. The cheapness of that land, in communities that have been disinvested for decades through documented policy choices, is being monetized by the most profitable industry in the world.
“A data center should not be a potential death sentence for a community’s health,” Abre’ Conner, Director of Environmental and Climate Justice at the NAACP, told theGrio earlier this year. Conner was speaking about a specific case, the NAACP’s suit over Amazon’s temporary diesel generator fleet at its Canton, Mississippi campus, in a county that is 36.4 percent Black. But her framing applies broadly. The top five hyperscale technology companies, Amazon, Google, Microsoft, Meta, and Oracle, have committed up to $690 billion in capital deployment for 2026 alone. Amazon alone is investing $25 billion across three Mississippi counties, each of them heavily Black. The investment carries job promises and grid upgrade pledges. But the structure of who bears the costs and who captures the value has not been fundamentally renegotiated from the historical norm.
The workforce dimension compounds the inequity. A hyperscale data center construction project employs hundreds of workers, laborers, electricians, ironworkers, pipefitters, over a multi-year build cycle. Those construction workers are predominantly non-union in many jurisdictions, concentrated in trades where Black representation lags in the high-wage specialty categories. When the facility opens, the permanent operations workforce, higher-paid network engineers, electrical systems technicians, data center managers, requires credentials concentrated in pipelines that underrepresent Black graduates. And at the AI company leadership level, Black representation in technical executive roles remains below five percent across most major firms.
The algorithmic dimension of the AI economy adds a layer that deserves direct confrontation. In 2018, MIT researcher Joy Buolamwini and Timnit Gebru published the Gender Shades study, documenting error rates of up to 34.7 percent in commercial facial recognition systems applied to darker-skinned women, compared to as low as 0.8 percent for lighter-skinned men. Eight years later, the pattern of algorithmic bias in consequential systems affecting Black Americans has been replicated and expanded. Healthcare algorithms recommending less aggressive care for Black patients at equivalent medical need. Hiring algorithms screening out Black-associated names. Predictive policing feedback loops concentrating surveillance in Black neighborhoods and generating self-reinforcing crime data. Each of these systems is trained on historical data that reflects historical discrimination and each deploys that discrimination at the scale and speed of computation.
“The AI industry’s default setting is the historical exclusion of Black people,” Charles Zackary King, host of America in Black and White, argued in tonight’s broadcast. “That is not a bug. It is an output. And an output is a design choice.” The point is not that bias is the intention of AI engineers. It is that the structural conditions under which AI systems are designed, who is in the room, whose data is in the training set, whose lived experience is treated as a sufficient test case, produce predictably inequitable outputs when the structural conditions are themselves inequitable.
The talent pipeline gap is the third leg of this stool. HBCU computer science and AI programs are growing, Howard’s Human-Centered AI Institute, Spelman’s ethical AI curriculum, NC A&T’s engineering programs, FAMU’s AI and Data Science Cluster, producing qualified graduates at increasing rates. But corporate AI recruiting pipelines remain concentrated at a small set of elite, predominantly white research universities. HBCU CS graduates are recruited into AI roles at a fraction of the rate of comparable non-HBCU graduates, not because of capability gaps, but because of structural recruiting gaps that major tech companies have the resources to close and, in most cases, have not.
The venture capital dimension is perhaps the starkest. Black-founded AI companies receive approximately one percent of the more than $40 billion annually deployed into AI startups. The structural barriers are documented: network effects in VC, pattern-matching that rewards familiarity over rigor, the near-absence of Black general partners at major firms. A growing ecosystem of Black-led VC funds and Black tech accelerators is working to close this gap. They are not yet operating at the scale the gap requires.
The picture that emerges from these convergent data points is not one of incidental or random inequality. It is a picture of structural exclusion operating at the precise moment when the architecture of the AI economy is being set. The decisions being made today, about where to site data centers, which communities to recruit from, which founders to fund, which systems to audit for bias, will shape the racial economics of the AI era for decades. The window to shape those decisions in a direction that produces equitable outcomes is narrow, and it is open right now.
The accountability begins with the AI industry itself: with data center developers who should not break ground in any community without a negotiated community benefit agreement; with AI companies whose recruiting pipelines should include HBCUs with the same depth and continuity as they include MIT and Stanford; with venture capital firms whose portfolio demographics should reflect a genuine commitment to identifying talent beyond their existing networks. It extends to policymakers at the federal, state, and local level, who have the regulatory authority to require bias auditing of consequential AI systems, to mandate community benefit agreements as a condition of public tax abatements, and to fund HBCU computing infrastructure as a national competitiveness priority. And it extends to Black communities, whose civic engagement with the data center going up in their district, the algorithm operating in their school or hospital or police department, and the venture ecosystem that does or does not fund their entrepreneurs, will determine whether this generation of economic transformation follows the historical pattern or begins to bend away from it.
The machine is being built. The question is for whom.
Please watch live October 19th at 6pm EST by clicking the YouTube Link below:
