Introduction
A plantation is not primarily a place. It is a structure. The structure has six features that recur across every instance the word has named historically, from the sugar islands of the seventeenth century to the cotton South of the nineteenth to the company towns of the twentieth.
A plantation concentrates land, water, and energy under owners who do not live on the land they own. The owners are absentees by design. The distance protects them from the consequences of what is done in their name. The labor extracts value from the land. The owners extract value from the labor. The value moves outward to the centers where the owners live, never circulating within the place where it was produced.
A plantation operates with a racialized division of who extracts and who absorbs the costs. A plantation externalizes its harms onto the surrounding ecology. A plantation produces wealth that its owners convert into political power. A plantation depends on the invisibility of its workers and its costs. A plantation, when it can no longer extract, abandons the place it ruined and moves to a new one.
This report documents an artificial intelligence infrastructure build-out in the United States that meets every one of these six criteria.
The build-out concentrates water, electricity, and computational capacity in facilities owned by approximately six companies whose executives do not live in the communities being drained. The labor displacement falls disproportionately on Black, Hispanic, and women workers. The wealth produced flows to a concentration in the top one percent of households that has not been seen in the United States since 1929. The plantation as a structure is returning. This is not an analogy reaching for rhetorical force. It is a description reaching for accuracy.
Criterion 1
Concentrates land, water, and energy under absentee owners.
Criterion 2
Operates with a racialized division of who extracts and who absorbs the costs.
Criterion 3
Externalizes its harms onto the surrounding ecology.
Criterion 4
Converts wealth into political power.
Criterion 5
Depends on the invisibility of its workers and its costs.
Criterion 6
Abandons the place it ruined when it can no longer extract.
Section I
A plantation concentrates land, water, and energy under owners who do not live on the land they own.
Six companies hold approximately forty percent of the S&P 500 market capitalization: Microsoft, Apple, Nvidia, Alphabet, Amazon, and Meta. The executives of these companies do not live in the communities where their data centers are built. They do not drink the water their cooling towers consume. They do not pay the rate increases their interconnection requirements force onto residential ratepayers. They do not breathe the air thickened by the backup diesel generators their facilities require.
They live, instead, in concentrated geographic enclaves — Atherton, Hillsborough, Belvedere, the Hamptons, Aspen, Jackson Hole, Lanai. The land they own at scale is bought as exit infrastructure rather than as residence. The communities absorbing the costs of their build-out are by design unknown to them, and the design is the point.
Source: EPA, Lawrence Berkeley National Laboratory. Indirect consumption through power plant operations adds 211 billion gallons annually by 2028.
The siting decisions are made in corporate boardrooms and ratified by state economic development offices competing for the appearance of growth. The residents of the affected states are not asked. The interconnection costs are socialized onto residential ratepayers through utility rate cases that the public lacks both the technical capacity and the legal standing to contest meaningfully. The wealth produced by the facilities flows to shareholders concentrated in coastal urban centers and offshore tax jurisdictions, never returning to the places that pay for it.
40%
S&P 500 cap held by 6 companies
211B
Gallons/yr indirect water 2028
≈6
Companies driving the build-out
This is criterion 1. The data shows it is met.
Section II
A plantation operates with a racialized division of who extracts and who absorbs the costs.
The labor displacement underway through AI deployment is not race-neutral, gender-neutral, or class-neutral. The pattern of who is being displaced first, who is absorbing the wage compression, and who is being denied the replacement jobs follows the same demographic contours that every previous extractive transition in American history has followed.
Source: Chicago Federal Reserve, Brookings Hamilton Project. Black workers overrepresented in 17 of the 30 US occupations at highest automation risk. Hispanic workers in 13.
Source: Center for American Progress, Brookings Hamilton Project. Approximately 500,000 Black Americans are employed in customer support roles alone — the bulk of them women.
The replacement jobs being created are concentrated in technical roles requiring credentials that the displaced workers do not hold and cannot acquire on the timeline of their rent. The median pay for AI-related job openings in 2025 reached approximately $157,000, with AI engineer roles growing 143 percent year over year. The workers absorbing the displacement are not the workers being hired into the replacement positions. The wage that disappears from one demographic appears as compensation for another.
$157K
Median AI job opening pay, 2025
+143%
YoY growth, AI engineer roles
500K
Black Americans in customer support
▣ Historical Fit
In the cotton South, enslaved labor was the extraction. In the Appalachian coalfields, immigrant and Black labor was the extraction. In the Dust Bowl plains, tenant farmers and migrant laborers were the extraction. In the deindustrialized Rust Belt, unionized industrial labor was the extraction. In each instance, the racial and class position of the extracted workers was engineered to limit their political standing to refuse the extraction, and in each instance, the wealth produced flowed to an ownership class whose racial position made their accumulation appear natural.
This is criterion 2. The data shows it is met.
Section III
A plantation externalizes its harms onto the surrounding ecology.
The water consumed by data centers is consumptive use. It is evaporated through cooling towers and lost to the watershed. It does not return to the aquifer. The aquifers being drawn down were filled by ten thousand years of rain. The drawdown is occurring on a timescale of years.
The electricity consumed by data centers is approximately 56 percent fossil-fueled. The carbon emissions of US data centers reached 105 million metric tons in 2024, equivalent to roughly 2 percent of all US carbon emissions. Methane leakage from the gas infrastructure built to serve the load adds an additional radiative forcing burden that the IPCC inventory methods do not fully capture.
A Time Magazine investigation in August 2025 documented sharp increases in ambient nitrogen dioxide concentrations in communities adjacent to large data center installations. The fenceline communities are disproportionately poor, disproportionately Black and Hispanic, and disproportionately rural. None of these costs appear on the balance sheets of the six companies driving the build-out. The accounting design is the point.
56%
Data center electricity from fossil fuels
105M
Metric tons CO₂, US data centers 2024
≈2%
Share of all US carbon emissions
▣ Historical Fit
The cotton plantation exhausted the soil and the owners moved west. The Appalachian coal operation collapsed the mountains and the owners moved to Pittsburgh. The Dust Bowl exhausted the topsoil and the owners moved to California. The deindustrial collapse left contaminated industrial sites and the owners moved to Sun Belt suburbs. The contemporary data center build-out is drawing down aquifers and grids and the owners are moving to bunkers and rockets. The pattern is the same. The destination is new.
This is criterion 3. The data shows it is met.
Section IV
A plantation produces wealth that its owners convert into political power.
Source: Economic Policy Institute, BLS. The space between these two lines is the fifty-year transfer that built modern wealth concentration.
Source: Economic Policy Institute CEO Pay Tracker. Average CEO compensation at top 350 firms reached ~$23M in 2024, with 79% paid in stock-related equity.
The wealth produced by the build-out is being converted into political power at a pace that has overwhelmed the regulatory infrastructure designed to constrain it.
Lobbying expenditures by the top five technology companies exceeded 96 million dollars in 2024 — not including trade association spending, dark money flows through 501(c)(4) organizations, or the personal political contributions of the executives themselves. Industry-funded think tanks shape the policy conversation in Washington. Industry-purchased media properties shape the public conversation. Industry-aligned judges sit on the federal bench. The conversion of economic concentration into political power is not hidden. It is documented in the public record. It is also the mechanism that has prevented the regulatory response that the data in this report would normally trigger.
This is criterion 4. The data shows it is met.
Section V
A plantation depends on the invisibility of its workers and the invisibility of its costs to the broader public.
The data center is the architectural expression of this criterion. A hyperscale facility is a windowless rectangle on the scale of forty football fields, sited deliberately in industrial parks and rural acreage where it cannot be seen from any public road. The facilities do not display their corporate ownership. They do not announce their function. They are unmarked by design.
The Framing Switch
The marketing copy
AI is a technology.
The description
AI is a labor arrangement.
The marketing copy
AI is a productivity tool.
The description
AI is a wage compression device.
The marketing copy
AI is a democratizing force.
The description
AI is a wealth concentration mechanism.
Source: Challenger, Gray and Christmas; BLS. The gap reflects structural reluctance by employers to name AI as cause — protecting share prices, brand reputations, and executive compensation.
This is criterion 5. The data shows it is met.
Section VI
A plantation, when it can no longer extract, abandons the place it ruined and moves to a new one.
The owners of the AI build-out have been preparing their exit for at least a decade. The exit infrastructure is documented in the public record. It is named by the people building it. It is not paranoia to describe it. It is paranoia to ignore it.
Figure 6.1 — Exit Infrastructure of Named Billionaires
Peter Thiel
477 acres in New Zealand, citizenship obtained. Stated publicly that democracy and freedom are incompatible.
Mark Zuckerberg
$300M+ Hawaii compound with reported underground bunker construction. Kuleana family displacements ongoing.
Sam Altman
Big Sur land, IDF gas masks, guns, gold, potassium iodide — stated in same interviews expressing confidence AI will go well.
Larry Ellison
Owns 98% of Lanai. Proposed Oracle build a national surveillance network on behalf of the US government.
Elon Musk
SpaceX Mars colonization — framed as ambition, functions as exit. Does not solve what his earth businesses cause.
Jeff Bezos
Blue Origin O'Neill cylinders — earth reserved for residential use, heavy industry moved off-planet.
Reid Hoffman told the New Yorker in 2017 that more than half of the Silicon Valley billionaires he knows have purchased some form of apocalypse insurance. Douglas Rushkoff documented in his 2022 book Survival of the Richest a group of hedge fund executives who flew him out to ask one question: how to keep their security forces loyal after their money no longer meant anything.
The exit confirms what the rest of the report has documented. The extraction is intentional. The owners know what they are doing. The owners do not intend to stay to absorb the consequences.
This is criterion 6. The data shows it is met.
Section VII
The plantation as a structure is not new to the United States. It is one of the country's most reliable institutional forms. Every generation has produced a new instance, and every instance has followed the same six criteria. The current AI build-out is the latest. It will not be the last unless something interrupts the cycle.
Caribbean Sugar, 17th c.
European absentee owners. Enslaved African labor. Soil exhaustion. Atlantic political capture. Invisible to European public. Abandoned when soil depleted. Descendants of enslaved remain. Descendants of owners do not.
Cotton South, 19th c.
Concentrated land. Enslaved labor. Soil exhaustion. Federal capture through three-fifths apportionment. Invisible to Northern public until abolition forced visibility. Abandoned South to consequences.
Appalachian Coal, 1880s–1940s
Absentee corporate owners in Pittsburgh, New York, London. Immigrant and Black labor. Company-town control. Mountain and stream devastation. Captured state legislatures. Abandoned when coal seams thinned.
Dust Bowl, 1920s–30s
Absentee landowners and corporate farms. Tenant and migrant labor. Topsoil exhausted through monocropping. Agricultural states captured. Invisible until dust storms made it impossible. Abandoned.
Rust Belt, 1970s–80s
Corporate owners in New York, then offshore. Unionized industrial labor. Environmental costs on Rust Belt cities. Both parties captured. Abandoned for Sun Belt, then Mexico, then China.
AI Build-Out, 2020s–
Six tech companies. Black, Hispanic, women workers displaced first. Aquifer and grid extraction. Lobbying and regulatory capture at record pace. Workers and costs invisible by design. Owners building the exit.
Section VIII
The trajectories documented in this report extend forward to 2050 under three scenarios. The scenarios differ only in the assumption about whether structural intervention occurs and at what scale.
Continuation
No intervention
Partial Intervention
Moderate reform
Structural Intervention
Data-warranted action
Scenario C is not unrealistic. It is the scenario the data warrants. The gap between what the data shows is necessary and what the institutions are capable of producing is the central political problem of the next decade.
Section IX
The historical record on what survives a plantation when the plantation falls is consistent across all five previous American instances. The institutions that absorbed the displacement and held the people together were not the institutions that had produced the displacement. They were institutions that pre-existed the plantation, survived under it, and were available when the plantation ended.
Post-Emancipation South
The Black church. The mutual aid society. The Black college. The Black-owned business district. The extended kinship network.
Post-Appalachian Extraction
The union halls. The church basements. The volunteer fire companies. The regional folk traditions. The family farms that refused to sell mineral rights.
Dust Bowl Migration
The church congregations that reformed in California. The family networks pooling resources across state lines. The labor organizing that built the United Farm Workers.
Deindustrialized Midwest
The union pension funds. The community development financial institutions. The neighborhood organizations. The immigrant communities. The Black churches, mosques, and Latino parishes.
The plantation cannot reach the neighborhood because the neighborhood operates at a scale and through a logic that the plantation does not know how to engage. The plantation can extract from labor. The plantation cannot extract from love. The plantation can capture government. The plantation cannot capture the church basement. The plantation can buy media. The plantation cannot buy the conversation between two neighbors who have known each other for twenty years.
Section IX — What Holds
The recommendation of this report is structural. The recommendation is to invest, while there is time, in the institutions that have historically held when the larger structure has fallen. The mutual aid networks. The cooperatives. The community land trusts. The Black churches and the immigrant parishes and the union halls and the credit unions. The associational life of the neighborhoods that the algorithm cannot enter and the bunker owners cannot understand.
The plantation owners are building their exit because they know what they have done. The neighborhoods are building their staying because they know what is coming. Both knowledges are correct. Only one of them produces a future that the rest of us can live in.
Appendix — Methodology, Sources, and Citation
Methodology
Historical data through 2024 is drawn from named institutional sources. Where projections extend beyond 2024, they are extensions of trends established in the historical data. They are not forecasts published by the cited institutions. The plantation analogy is argued by structural criteria — each section names a criterion, presents the data, and concludes only after the data has been presented. Readers who reject the analogy are invited to identify which of the six criteria is not met and on what evidence.
Primary Sources
Federal Reserve Survey of Consumer Finances · Economic Policy Institute, CEO Pay Tracker and Productivity-Pay Gap analysis · DOE Lawrence Berkeley National Laboratory, 2024 US Data Center Energy Usage Report · EPA data center water consumption analysis 2025 · Electric Power Research Institute · Brookings Institution Hamilton Project · Chicago Federal Reserve Working Paper 2024-03 · Center for American Progress · Challenger, Gray and Christmas monthly layoff reports · Gartner AI workforce projections 2025 · New Yorker, Wired, Forbes, Bloomberg, Reuters, WSJ, NZ Herald 2012–2026 · Douglas Rushkoff, Survival of the Richest (2022)
Intellectual Debts
Cedric Robinson on racial capitalism · Saidiya Hartman on the afterlife of slavery · Ruha Benjamin on the New Jim Code · Safiya Noble on algorithms of oppression · Mary Gray on ghost work · Astra Taylor on democracy and platforms · Ivan Illich on tools for conviviality · Naomi Klein on disaster capitalism · John McKnight on asset-based community development
Suggested Citation
McAleavey, Marc. (2026). The Plantation Returns: Closed Circuits, Tech Billionaires, and the Illusion of Democratic Consent in the AI Economy. Spatial Liberty Report No. 1. Joy Repair / Meta-Macro Social Work. Indianapolis, IN.
Published under Creative Commons Attribution-NonCommercial-ShareAlike. Republication and adaptation permitted with attribution. Contact: Marc McAleavey, MSW · Joy Repair · Indianapolis, Indiana · joy4repair.com
The Plantation Returns · Spatial Liberty Report No. 1 · Joy Repair · May 2026
The neighborhood is not the constituency for the complex. The neighborhood is the complex.