Prefer to read offline? The same corpus is available as a curated reference PDF, built for the Sloan Foundation research team at Emory and the University of Tennessee.
Download the PDF.
Every claim is built from primary documents; each volume is deposited and SHA-256 hashed on Zenodo, and most carry an SSRN twin for indexing. Five entries are marked ★ Anchor — the recommended starting set for a reader new to the corpus.
Coverage: Statewide, current through July 25, 2026 · v5.3 · v-next in preparation (landscape actively developing)
Evidence: Statewide inventory of 286 facilities; JLARC 2024 study, University of Georgia econometric model, Dominion Virginia Power 2025 filings
Length: Reference volume, extended
Argument. Virginia's data-center sales-and-use tax exemption cost the Commonwealth $1.94 billion in forgone FY2025 revenue; the buildout exposes residential ratepayers to costs Dominion's own filings project will roughly double monthly bills within a decade.
Evidence. Statewide inventory of 286 facilities; JLARC's December 2024 study, University of Georgia econometric modeling, and Dominion Virginia Power's 2025 audited filings (data centers at 28% of Virginia Power sales; ~25% of Virginia electricity overall).
Cite for. The fiscal-exemption estimate, the statewide inventory record, and the ratepayer-exposure argument. The reference volume for the entire county case-study series.
Companion. Read alongside Governing the Demand Shock (Report) and (Book Edition) — the reference gives the totals; the Report gives the mechanism.
★ ANCHORGoverning the Demand Shock: Local Institutional Variation in Virginia's Data Center Expansion (Report)
Coverage: Virginia, 2018–2026 · v-next in preparation (landscape actively developing)
Evidence: 13 jurisdiction-level case studies; statewide inventory of 286 facilities
Length: Analytical report
Argument. The decisive variable separating communities that admitted the industry on terms from those that did not was neither geography nor project size — it was the timing of local institutional reform relative to the industry's arrival, and behind timing, the leverage a community still held when it acted.
Evidence. Thirteen jurisdiction-level case studies with the statewide inventory of 286 facility records. Develops a five-posture spectrum: reform-after, refusal-before, contract-lock, governance-capture, and pre-entitlement-by-ordinance.
Cite for. The leverage-and-timing framework, the five-posture spectrum, and the geographic diffusion analysis. The most compact academic-format statement of the argument that runs across the county case studies.
Companion. Companion to the Commonwealth Reference (20679292). Book Edition (20697583) extends to a general audience.
Governing the Demand Shock: Data Centers, Local Power, and the Distribution of Cost in Virginia (Book Edition)
Argument. One industry, one state subsidy, one utility, the same eight-year window, opposite distributions of cost across similarly situated jurisdictions — treats announced investment as a promotional genre rather than an accounting category.
Evidence. Grounded in a fifteen-year Mecklenburg audit and the Fredericksburg fall-line natural experiment. Opens on Hanover turning away a 900-MW campus on water while forty miles south Chesterfield approved its largest campus by right.
Cite for. The audit-vs-promise distinction, the Hanover/Chesterfield paired case, and any need for a public-facing narrative treatment.
Companion. Sibling to Governing the Demand Shock (Report). Use the Report for compact academic citation, the Book for narrative accessibility.
II. Utilities and Generation (RQ2)
Investor-owned, cooperative, merchant, and federal utility regimes; transmission cost allocation; behind-the-meter generation; construction-trades labor.
★ ANCHORThirteen Cooperatives: Virginia's Distribution Cooperatives and the Arrival of Load Larger Than the System
Coverage: Virginia, 2018–2026, with statutory analysis through July 2025
Evidence: All 13 VA distribution cooperatives; 4 large-load rate schedule cases; 7-case affiliate authority sequence; SCC dockets, ODEC filings
Length: Reference volume
Argument. Thirteen electric distribution cooperatives serve Virginia — the figure most often cited is eleven. Data-center load is arriving on several of them, and the SCC regulates cooperatives on terms that differ from the investor-owned utilities in ways the public record rarely sets out in one place.
Evidence. Walks identity and jurisdiction, wholesale supply (9 of 13 through Old Dominion Electric Cooperative), the seven-case affiliate sequence in which one cooperative was twice refused authority to serve hyperscale load through an unregulated affiliate, and the four large-load rate schedule cases read side by side against the investor-owned regime. Addresses Va. Code § 56-234 D (2025 Acts of Assembly, chs. 499 and 598).
Cite for. The cooperatives-vs-investor-owned regulatory distinction, the ODEC wholesale structure, the affiliate-authority sequence, and the statutory reframe. Directly fills the proposal's stated gap on cooperatives in data-center deals.
Evidence: 108 primary documents, each hashed; four generation routes coded; SCC dockets, VCEA filings, DEQ air-permit register
Length: Working paper, evidence-graded
Argument. Virginia has one statute governing generation-facility construction. Four routes now serve data centers; the statute reaches two, applies its full test to one, and does not reach the other two at all.
Evidence. 108 primary documents, each SHA-256 hashed. VCEA reliability-exception plant drew 1,923 comments, 83 witnesses, 4 days of hearing, 338 filings; merchant plant reviewed under narrower two-prong test; 132-MW behind-the-meter gas plant running since April 2024 with no certificate; ~8,000 permitted backup generators outside SCC dockets entirely.
Cite for. The four-route framework, the utility/merchant/behind-the-meter/backup taxonomy, and the load-forecast case where a utility acted on class-of-forecast and the load did not appear.
Companion. Source manifest at 22099427 is the reproducibility apparatus. Sits alongside Google–CFS PPA (20679935) as the two generation-regime pieces.
Source Manifest and Datasets for Built for Someone Else
Argument. The reproducibility apparatus for Built for Someone Else. A researcher can re-fetch every URL, hash the bytes, and confirm whether they are reading what the volume read; a mismatch flags subsequent publisher revision.
Evidence. 125 files SHA-256 hashed; for 101 of the 108 primary source documents the URL of origin is recorded. RETRIEVAL_NOTES.md covers documents whose links won't last (SCC opaque tokens, 2020 legislative record on a legacy system, FERC eLibrary session-only). Five datasets with blind-recode second-coder sheets, including the DEQ air-permit register (201 issued, 190 facilities, 37 pending).
Cite for. Reproducibility apparatus for 22096613. Also useful independently for the DEQ air-permit register.
Companion. Parent volume: Built for Someone Else (22096613).
★ ANCHORConcentrated Benefit, Dispersed Burden: Who Carries Virginia's Data Center Transmission (Volume I)
Argument. The benefit of a data center is captured almost entirely by the locality hosting it, the cost of the wires is spread across an entire utility zone with no reference to geography, and the physical burden falls on a third set of places that appear in neither calculation.
Evidence. PJM allocates the Joshua Falls–Yeat 765 kV project under a mandatory two-part formula for facilities ≥500 kV: half by load-ratio share, half by distribution factor. Dominion's combined share is 44.96% as filed February 2025, 45.29% on the 2026 vector. Against published route geometry, 80.99% of the routed line runs through counties with no operating data center; benchmarked against 2,000 random 116.47-mile corridors in Virginia.
Cite for. The FERC cost-allocation record, the geography-blindness of the tariff, and the random-corridor test.
Companion. Vol II (21996375) extends to TVA. Companion dataset: Transmission Siting Disclosure Capture (21959988).
The Whole Record: What the Tennessee Valley Authority Publishes When Nothing Requires It (Volume II)
Argument. TVA builds transmission across seven states and files no siting case for any of them. The GAO reported the condition to Congress in 1982 and proposed statutory fixes; none were enacted. What TVA chooses to publish is the public record — this volume measures it.
Evidence. Eighteen TVA transmission project pages coded field-by-field (16 name no end-use customer or load type; no page states a megawatt figure; environmental-review section identical across all 18). TVA studies 2.63 miles of corridor per selected mile. Categorical-exclusion record reconstructed from 11,899 published determinations.
Cite for. TVA's absent siting-review record, the 1982 GAO precedent, and the categorical-exclusion arithmetic. The federal-utility parallel to Vol I.
Argument. Fixes bytes as of retrieval so findings drawn from live utility ArcGIS services remain checkable when the source revises.
Evidence. TVA route alternatives (2,645 records), selected centerlines (447), rights-of-way, substations, service definitions; the complete 302-service TVA organization directory. Oncor/LCRA Permian CCN and Dominion Energy Virginia service captures for comparison.
Cite for. Versioned record when a subsequent utility change alters what the live service displays.
Companion. Companion to Concentrated Benefit Vol I (21895588) and Vol II (21996375).
When the Operator Builds the Plant: The Google–CFS Power Purchase Agreement
Coverage: Chesterfield County, VA — agreements dated 2025
Evidence: Google 200 MW PPA (June 30, 2025); Eni offtake (Sept 22, 2025, >$1B); equity stakes; SCC dockets
Length: Case study
Argument. Documents the commercial instruments between Google, Eni, and Commonwealth Fusion Systems for the 400-MW ARC fusion plant in Chesterfield County (Fall Line Fusion Power Station). Not a technology profile; examines the commercial instruments themselves.
Evidence. Google's 200 MW agreement of June 30, 2025; Eni offtake of September 22, 2025 valued by parties at >$1 billion; equity stakes both buyers hold in the developer. Treats the deals as a mechanism for buying down technology risk with reputational and balance-sheet capital.
Cite for. The operator-anchored generation model, the specific commercial terms, and the risk-transfer analysis.
Companion. Chesterfield County case study (20672436). Thirteen Cooperatives (22651576) for the broader utility-regulatory context.
Half of a 765-kV Line in PJM is a Postage Stamp; the Other Half Stops at the Zone Line
Evidence: Same primary-document base as Concentrated Benefit Vol I
Length: Trade-press piece
Argument. Public-facing companion to Concentrated Benefit Vol I. Argues PJM's Schedule 12 allocates half the cost of a data-center-driven transmission line by geography and half by ignoring it.
Evidence. Reads Schedule 12 and the Joshua Falls–Yeat allocation for the RTO Insider trade audience (utilities, ISO/RTO staff, FERC practitioners, state regulators, transmission developers).
Cite for. Trade-press summary of the Vol I argument in a venue read by the actors making these decisions.
Companion. Concentrated Benefit Vol I (21895588).
Where the electricians went: Hyperscale data center construction and building trades membership in federal union filings, 2018 to 2025
Evidence: 89 metropolitan IBEW locals; 8-year LM-2 panel; QCEW corroboration for 86 locals; hyperscale campus published by 4 operators
Length: Working paper v0.1, with complete replication package
Argument. Published employment figures for data-center construction come from the industry, from agencies relaying industry figures, or from unions testifying on projects. Builds an independent series from federal filings.
Evidence. Federal LM-2 filings (under penalty of perjury) joined with IBEW jurisdiction maps. Eight-year panel of inside electrical construction membership for 89 metropolitan locals. Median treated local grew 33.7% between 2018 and 2025 against 13.2% elsewhere; regression difference 25.8 percentage points, randomization p < 0.001. Event study divergence significant from 2023. QCEW corroboration for 86 locals.
Cite for. The federal-filings labor measurement, the hyperscale-treated growth differential, and the replication apparatus. Directly fills the proposal's stated gap on who staffs data centers and worker justice.
Companion. Stands alone in the corpus. Mecklenburg County (20673484) is the closest county-level treatment of the labor-promise-vs-actual question.
III. Water: A Scope-Adjacent Thread
Water is outside the proposal's stated scope but the record has implications for RQ3 justice framing and for the Virginia–Georgia–Tennessee cross-state comparison built into RQ2.
Two Systems, One Drought: Virginia's Groundwater Ceiling, Its Rivers, and the Water the Energy Buildout Actually Uses
Coverage: Virginia, calendar year 2024 basis; recent-record period through March 2025 · v1.3
Evidence: DEQ 2025 Annual Water Resources Report (RD512, CY 2024); Fairfax Water 2024 withdrawal record; Lake Moomaw operational record
Length: Analytical report, 218 endnotes
Argument. Documents the two water regimes data-center load in Virginia actually touches: the Coastal Plain groundwater system east of the fall line, and the surface-water regime including the James River and the reservoirs feeding Fairfax Water.
Evidence. Fairfax Water's two treatment plants withdrew a combined 164.0 MGD in 2024, both unpermitted under the pre-1989 grandfather exclusion. Two Lake Moomaw release modifications on the recent record (December 2023, December 2024 maintained through March 2025). v1.3 (26 August 2026) completes an edition migration; 218 endnotes, zero orphans.
Cite for. The two-regime framing, the Fairfax Water figures, and the Lake Moomaw curtailment record.
Companion. Reads with Aquifer Says No (21556435) and Line Drawn on Highway (22105181). Where the Instrument Stops (22116053) is the compact synthesis of the three.
The Aquifer Says No: DEQ's SJ 25 Groundwater Study and the Data Center Question East of I-95
Argument. A data center on evaporative cooling, or any comparable water user, is unlikely to find a reliable, sufficient groundwater supply anywhere in the Virginia coastal plain.
Evidence. DEQ's nine-site technical evaluation simulated a hyperscale-scale withdrawal (3 MGD, evaporative cooling) across the Virginia Coastal Plain aquifer system; all nine locations failed the permitting criteria. Best case 360,000 gpd at Surprise Hill in Northumberland; near the I-95 corridor maxima fall to 60,000 gpd and below; Henrico and Sussex test sites failed at any measured volume.
Cite for. The DEQ nine-site test, the "aquifer says no" summary of SJ 25, and the study's delay/release history (six months past statutory deadline).
Companion. Two Systems (21816008), Line Drawn on Highway (22105181), Where the Instrument Stops (22116053) — same water thread.
A Line Drawn on a Highway: Virginia's Groundwater Boundary, the Rivers Beyond It, and the Data Center Load the Regime Does Not Reach
Argument. Virginia regulates groundwater by asking three categorical questions and never asking how much. Data centers answer all three in the way that exempts them.
Evidence. The groundwater program asks where — inside two declared management areas it regulates; outside them not at all. The surface-water program asks when — pre-1989 withdrawals need no permit; neither of Fairfax Water's two raw water intakes holds one. Statewide, 78% of reported non-power withdrawals are unpermitted. Of 286 inventoried data-center facilities, 257 (89.9%) sit outside every declared groundwater management area; every one of Loudoun's 121 facilities sits north of where I-95 leaves Virginia.
Cite for. The three-question exemption structure, the 89.9% coverage figure, and the I-95-as-boundary finding.
Companion. Reads with Two Systems (21816008) and Aquifer Says No (21556435). Executive brief: Where the Instrument Stops (22116053).
Where the Instrument Stops: Virginia's Groundwater Ceiling, Its Grandfathered Rivers, and the Load Outside Both
Coverage: Virginia, synthesis of prior three water volumes
Evidence: Cross-volume analysis of 726 entity-bearing claims across three volumes
Length: Executive brief
Argument. Executive-brief synthesis of the three deposited Virginia water volumes, organized on the three functions of SJ 25.
Evidence. Across 726 entity-bearing claims in the three volumes, four subjects carrying twenty or more claims are engaged near-evenly by all three. DEQ tested nine sites and none passed at 3 MGD. Of 286 inventoried data-center facilities, 257 sit outside every declared groundwater management area; in the northern crescent the nearest facility to I-95 is 12.7 km away.
Cite for. A compact synthesis of the water thread; use when the individual volumes are too long for the audience.
Companion. Synthesizes 21556435, 21816008, 22105181. Cross-state extension: No Field to Land In (22390449).
Declared, Not Measured: Volumetric Permit Triggers and Sub-Threshold Industrial Water Demand in Georgia
Coverage: Georgia, statutory analysis through July 2026
Evidence: Ground Water Use Act + 16 implementing rules; 16 DRI filings January 6 – July 22, 2026
Length: Working paper
Argument. Georgia's 100,000-gallon-per-day groundwater permit trigger runs per person and turns on a quantity that, before a permit issues, exists only as the withdrawer's declaration. A measurement-coincident threshold — the regime regulates what it can already see.
Evidence. Reads the Ground Water Use Act and its sixteen implementing rules provision by provision. Sixteen data-center Development of Regional Impact filings were made between January 6 and July 22, 2026. Stated independently as a limiting case of Bennear (2008).
Cite for. The measurement-coincident threshold framework, the Georgia sixteen-DRI docket, and the limiting-case generalization.
Companion. Extends the Virginia water thread; references 21816008 and 22105181. Cross-state extension: No Field to Land In (22390449).
No Field to Land In: How Three States' Water Withdrawal Forms Make Utility-Supplied Data Centers Illegible
Coverage: Virginia, Tennessee, Georgia — form analysis current September 2026
Evidence: Virginia OWS-3N; Tennessee CN-1119; Georgia's two monthly withdrawal forms; federal glossary
Length: Working paper
Argument. A data center that buys treated water from a public utility is, in Virginia and Tennessee by express clause and in Georgia by definition, not a person who withdraws water. That much is already public. What follows from it has not been examined: there is no field on any state's withdrawal-reporting form built to record the delivery.
Evidence. Virginia's OWS-3N defines a transfer as a transaction between two waterworks. Tennessee's CN-1119 asks a seller nothing about its buyers. Georgia's two monthly forms alone ask a supplier to identify its buyer, and what they ask for is a public water system by identifier or name. A facility that buys its water is not one and holds none.
Cite for. The cross-state form-field analysis and the record-void argument.
Companion. Reads with all four Virginia water volumes and Declared Not Measured (22240696).
IV. Local Governance and Stakeholder Response (RQ1)
Twelve Virginia jurisdictions, each a case study in how the industry moved through local government and what the community response looked like.
Case Study: Henrico County — How Two Neighborhoods Forced Virginia's Second-Largest Data Center Jurisdiction to Change Course
Argument. The template case study. Henrico's regulatory reversal was produced by an unusual political convergence: a predominantly Black, middle-income eastern community framing opposition as environmental justice, and a predominantly White, high-income West End community framing it as property-value protection, arriving simultaneously at the same structural demand.
Evidence. Documents the arc: 1990s $44M semiconductor bet; 2009 Qimonda collapse; 2010 QTS retrofit; 2017 tax cut that brought Meta; the by-right approval framework; the thirteen-month resistance arc that ended the framework June 10, 2025; Centra Logistics vested-rights denial (Feb 2026); May 2026 disclosure of a grandfathered QTS pipeline (17 buildings, ~8M sf) the new ordinance cannot reach.
Cite for. The template every subsequent county case is compared against. The Black/White cross-class coalition finding is the strongest environmental-justice case in the series.
Companion. Chesterfield (20672436), Hanover (20672048) — three central Virginia jurisdictions with divergent postures.
Case Study: Hanover County — Mountain Road, Tract, and the Search for Leverage
Argument. On May 27, 2026, the Hanover Board rejected Tract's Mountain Road Technology Park 4–3 — 430 acres, 900 megawatts, the largest single land-use decision the county had faced. The county that built the planning framework inviting the proposal refused the flagship application that framework produced.
Evidence. Twelve-year construction of the recruitment apparatus: EDZ Route 33 (2013), Strategic Zoning Initiative (2015), $0.45 equipment tax rate (2023), Small Area Plan naming data centers a "preferred business type" (Aug 2025). Three proposals resolved three ways: Hickory Hill (1,211 ac) approved unanimously 2024; Hunting Hawk withdrawn under unanimous PC opposition Jan 2026; Mountain Road rejected despite proffer cash growing from $90,000 to ~$21M. Water decided it.
Cite for. The leverage-window closing, the proffer-inflation-under-pressure pattern, and the water-as-decision-variable finding.
Companion. Chesterfield (20672436) — the paired case in Governing Demand Shock Book Edition (20697583).
Case Study: Chesterfield County — The $0.24 Rate, the Thirty-Year Lock, and the Two-Corridor County
Argument. Chesterfield built its data-center strategy on a single number. April 2019: equipment tax rate set at $0.24/$100 — a deliberate undercut of Henrico's $0.40 and the lowest in Virginia. The rate is locked for thirty years.
Evidence. Google assembled ~1,640 acres through shell entities; Chirisa committed $2B to Meadowville; Commonwealth Fusion Systems selected Battery Brooke for the world's first grid-scale commercial fusion plant. v3.1 corrects Rider CERC residential impact (SCC Final Order in PUR-2025-00037: $0.60/month at 1,000 kWh, rising to $1.14/month in 2027) and PJM zone-price figures (Dominion zone rose from $28.92 to $444.26/MW-day, a 1,436% increase).
Cite for. The tax-rate competition mechanism, the thirty-year lock as institutional design, and the CERC/PJM ratepayer figures.
Companion. Google–CFS PPA (20679935) covers the Battery Brooke fusion deal. Hanover (20672048) is the paired jurisdiction.
Case Study: Mecklenburg County — Boydton, South Hill, and What Fifteen Years of Data Centers Actually Bought
Evidence: 15-year fiscal ledger; employment records; township demographic time series; Microsoft public statements
Length: Case study — the audit case
Argument. The only Virginia jurisdiction where the promises are old enough to audit. Fifteen years after Microsoft's Boydton announcement, the audit returns a split verdict.
Evidence. Microsoft announced Boydton August 27, 2010 with $499M and 50 jobs pledged. Fifteen years later: eleven of the county's twelve data-center sites (~3,762 acres), just under $2B announced (public statements stopped 2019), ~700 jobs with vendors in a county of 30,000. Fiscal promise kept and exceeded — data-center taxes supply 35–40% of general-fund revenue, financed a $154M school complex without debt, drove real-estate tax to $0.31. Labor promise kept only in its narrowest reading. Boydton town shrank from 431 to 302 residents.
Cite for. The audit-vs-promise methodology and the fifteen-year fiscal record. The reference point for every prospective-analysis case.
Companion. Governing Demand Shock Book Edition (20697583) grounds its opening in this audit.
Case Study: The Fredericksburg Ring — Five Jurisdictions, One Corridor, and the I-95 Path South
Argument. The Fredericksburg ring is where the Northern Virginia diffusion wave crosses the river. Since 2018 the five have held themselves to an identical $1.25/$100 tax rate on data-center equipment — a regional compact to prevent shopping. The compact held. Everything else diverged.
Evidence. Since 2022, the five jurisdictions have approved data-center campuses totaling at least 5.4 GW of reported capacity and ~44M sf of entitled building area. Spotsylvania reversed its own two-month-old ordinance in February 2026. Stafford defended by-right March 2025, then adopted Virginia's most stringent setback that October and grandfathered all five approved campuses six weeks later.
Cite for. The regional-compact mechanism, the ring aggregation across five jurisdictions, and the divergent local reforms atop a shared tax floor.
Companion. Loudoun Endgame (22374193) is the northern parallel; both are diffusion-wave analyses.
Case Study: Orange County and the Town of Orange — Convergence at Two Scales
Coverage: Orange County + Town of Orange, February–April 2026
Evidence: Town and county ordinances; four county resolutions on the 765-kV line; Wilderness Crossing rezoning record
Length: Case study
Argument. Two governments whose chambers sit blocks apart regulated the same industry in the same season with different instruments — the Town removed data centers as by-right February 17, 2026; the County followed April 28, 2026 by confining the use to a floating Technology District that exists on no parcel of land.
Evidence. Town of Orange (~5,000): 5-member council removed data centers as by-right, added SUP gate. Orange County (~36,000): Board voted 5–0 to strike "data center" from the by-right industrial table. Same county spent the spring filing four resolutions, two sets of federal comments, and a five-county joint motion against the 765-kV line. None of this reaches the 2,602-acre Wilderness Crossing rezoning of April 2023 (732 acres for data centers, projected 1,500 MW).
Cite for. The two-scale convergent-reform pattern and the inability-to-reach-prior-entitlement finding.
Companion. Sits with the diffusion-arc cases: Louisa (20673455), Goochland (20673467).
Case Study: Louisa County — The First Overlay, the Amazon Retreat, and the $17 Billion Bet
Evidence: First Technology Overlay District record; AWS third-campus withdrawal; EdgeCore Shannon Hill announcement
Length: Case study
Argument. Louisa invented the instrument the rest of central Virginia is now copying and then walked away from its central feature. The same ordinance produced a hyperscaler that retreated from one fight and a developer that committed $17 billion, because the outcomes were decided by variables the instrument does not control.
Evidence. April 2023: Virginia's first Technology Overlay District, pre-entitling data centers by-right across ~6,400 acres. Within five months: $11B AWS commitment. August 2024: correction converts every use to CUP. July 2025: AWS withdrew its third-campus application under organized resident resistance — the only hyperscaler retreat documented in the series. June 2025: sale of the county-built Shannon Hill park to EdgeCore, which announced a $17B, 1.1-GW campus.
Cite for. The first-overlay history, the AWS retreat, and the instrument-vs-outcome finding.
Companion. Goochland (20673467) is the template's second Virginia adoption; Culpeper (20673529) preceded the codified overlay.
Case Study: Goochland County — The Overlay, the Lawsuit, and the Template
Argument. Every prior county in this series decided data centers project by project. Goochland tried something different — and residents sued twenty-eight days later.
Evidence. November 6, 2025, after a hearing past midnight: Board voted 4–1 to adopt a Technology Overlay District. ~4,400 acres pre-zoned for data centers, by-right in most of the district, incentives layered on top, use table including SMRs by conditional use permit. No speaker at the hearing supported it. Twenty-eight days later four residents sued; county has budgeted $250,000 to defend. Circuit court rejected the county's attempt to file all 8,500 pages of its adoption record; demurrer argued October 20, 2026. Tract filed the district's first application June 10, 2026 — a twelve-building campus.
Cite for. The pre-entitlement-by-ordinance posture, the notice-and-adoption litigation, and the template's spread.
Companion. Louisa (20673455) is the original template. Botetourt (20673512) documents a parallel concealed-recruitment mode.
Case Study: Botetourt County — Project Raspberry, the Water Authority, and the Wave That Crossed the Blue Ridge
Coverage: Botetourt County, 2024–2026 · v1.1, current through September 8, 2026
Evidence: Performance agreement text; option-signing timeline; joint PC/BOS session transcripts; Western Virginia Water Authority record
Length: Case study
Argument. The performance agreement obligates the county to help Google resist disclosure requests and assigns Google control of "all media and press relation services" concerning its involvement with the county — a concealed-recruitment sequence unmatched elsewhere in the series.
Evidence. June 24, 2025: Botetourt announced Google had bought 312.55 acres of the county's only business park for $14,055,406.37 plus $4M for county initiatives. Option signed October 1, 2024. Eight weeks after signing, the county added "data center" to permitted uses in the only zoning district that could host one, describing the change as housekeeping. Staff report: "As this is a speculative request, no use or concept is identified at this time." Google had been visiting since January 2024 under the code name Project Raspberry.
Cite for. The concealed-recruitment pattern and the FOIA-resistance obligation in a public agreement — a case study for Alaina's regulatory-capture / obstruction dissertation frame.
Companion. Prince William (20673498) is the other governance-capture case. Union Hill precedent cited (Fourth Circuit 2020).
Evidence: ~12–13M sf built or approved across 8 campuses; 2,435 MW of delivery-point requests (Dominion May 2026 count); McDevitt Drive consolidation record
Length: Case study
Argument. The quiet is an artifact of institutional design. In 2021 the county consolidated five scattered technology zones into a single 690-acre incentive district along McDevitt Drive and made data centers by-right. By-right needs no rezoning or hearing, so it generates no visible politics.
Evidence. One of the largest accumulations of data-center entitlements in Virginia's diffusion ring assembled without a Digital Gateway, a Mountain Road, or a marquee lawsuit anyone outside the county can name. Federal Reserve had buried its currency reserve in Mount Pony along the same corridor. Five campuses filled the zone (CloudHQ, DataBank, STACK, EdgeCore, Red Ace) without a single contested hearing. Where discretionary votes were required, Culpeper fought like everywhere else.
Cite for. The by-right-as-quiet mechanism and the McDevitt Drive consolidation as instrument design.
Companion. Louisa (20673455) and Goochland (20673467) — the pre-entitlement family.
Case Study: Prince William County — Governance Capture, the Revolving Door, and the Democratic Correction
Coverage: Prince William County, through July 2026 · v3.1
Evidence: Conflict-of-interest ruling; Supreme Court of Virginia filing record; primary/general election results; SCC PJM correction
Length: Case study — the governance-capture case
Argument. The case where governance itself was the failure point. The largest data-center proposal in American history was filed at the Supreme Court of Virginia three hours before deadline, then withdrawn on July 2, 2026. No court ever reached its merits.
Evidence. A sitting supervisor signed a contract to sell his own land to one of the developers ten days before joining the application. The board chair held a disclosed position of $50,001–$250,000 in the parent company of the other developer while the project moved through approval. The supervisor resigned under a conflict-of-interest ruling and joined a data-center lobbying firm within three months. The chair lost her primary. v3.1 (24 August 2026) corrects PJM Dominion-zone capacity-price figures.
Cite for. The governance-capture posture, the elected-official conflict-of-interest chain, and the ballot-box correction sequence. Direct match to Alaina Kinol's power-and-obstruction research frame.
Companion. Botetourt (20673512) is the other concealed-recruitment case. Loudoun Endgame (22374193) is the adjacent-county contrast.
★ ANCHORCase Study: Loudoun County, Virginia — The Endgame Case: What It Costs to Change the Rules After the Industry Is Load-Bearing
Coverage: Loudoun County, adoption March 18, 2025 · analysis through September 2026
Evidence: 7–2 board vote record; 5–3–1 Planning Commission denial recommendation; point-geometry join of built and under-construction footprint
Length: Case study — the diagnostic case
Argument. On March 18, 2025, the Loudoun Board voted 7–2 to require a special exception before a data center could be built in the districts where the use had been permitted by right. Adopted over its own Planning Commission's denial recommendation. It never reached the Route 28 corridor, where commercial and industrial zoning is frozen by state statute until 2037.
Evidence. A point-geometry join puts 51% of Loudoun's built data-center square footage inside that district, 43.16% of built and under-construction, and 26.23% of what is under construction alone — a constraint the county publishes on its own website but that went unmentioned across fourteen months of rewriting the ordinance and the client alerts and advocacy tracking around it.
Cite for. The load-bearing-industry reform limit, the Route 28 statutory-freeze exemption, and the diagnostic county at the top of the Virginia diffusion series.
Companion. Prince William (20673498) is the adjacent-county contrast. Louisa (20673455) is the first-overlay counterexample.
V. Cross-State Comparison (VA / GA / TN)
The two Deep-South statewide reference reports that parallel the Virginia work in Georgia and Tennessee.
What Georgia Is Spending, What Georgia Is Getting: A Statewide Data Center Report
Evidence: O.C.G.A. § 48-8-3(68.1); Carl Vinson Institute evaluation; Georgia PSC certification record; six residential rate increases
Length: Statewide reference report
Argument. Georgia built the second-largest data-center market in the US without ever deciding to. No statewide siting law; no statewide water or power assessment; the one bill to pause the central subsidy passed both chambers in 2024 and was vetoed. With the state silent, local government became the siting authority by default.
Evidence. Sales-and-use tax exemption waived ~$450.6M in 2024, rising toward ~$866.7M by 2030. Carl Vinson Institute found net fiscal impact negative every year and revised the share of activity actually caused by the exemption from 90% down to 30%. Georgia Power's new-generation need rose from ~400 MW (2022) to 8,500 MW (2025); PSC certified ~9,985 MW December 2025, ~80% tied to data-center demand and mostly gas. Residential customers absorbed six rate increases totaling ~$43/month. By June 2026, 32 counties and 21 cities had adopted or drafted data-center ordinances or moratoria.
Cite for. The Georgia fiscal record, the Carl Vinson revision (this is a load-bearing citation for the ratepayer-subsidy argument), and the local-siting-by-default mechanism.
Companion. Declared Not Measured (22240696) and No Field to Land In (22390449) are the Georgia-water pieces. Tennessee report (20717386) is the state-comparison counterpart.
What Tennessee Is Building, and What It Is Not Regulating: A Statewide Data Center Report
Evidence: TVA structure; 153 local distribution companies; 2026 gas-generation deregulation legislation; region-by-region facility census
Length: Statewide reference report
Argument. The utility serving Tennessee's data centers answers to no state regulator. TVA is a federal corporation with a presidentially appointed board and no state commission above it, selling power through 153 local companies that did not negotiate the loads landing on their systems.
Evidence. By mid-2026: roughly 61 operating or under-construction data centers across at least ten markets. Reconstructs the structural frame, the fiscal and policy record (incentive architecture, undisclosed cost, 2026 session that deregulated private gas generation, and the local turn), and a region-by-region facility census — from the xAI environmental-justice fight in Memphis to the crypto mines of the northeast mountains.
Cite for. The TVA-unaccountability finding, the 2026 gas deregulation, and the region-by-region facility census.
Companion. The Whole Record Vol II (21996375) is the transmission-specific companion. Georgia report (20717372) is the Deep South comparison.
VI. Supporting Datasets and Documentation
Research infrastructure supporting the corpus — the AI-assisted civic research methodology and the university-lab partnership proposal that underlies the primary-document work in the county case studies.
Regional Government Records Corpus: Institutional Documentation Package (Volumes I–III)
Evidence: Corpus overview; AI-assisted methodology; partnership proposal
Length: Three-volume package
Argument. Three volumes documenting the Regional Government Records Corpus: overview, AI-assisted civic research methodology, and partnership proposal for a Virginia Local Government Public Records Lab under a university roof.
Evidence. Vol I: holdings, architecture, cross-jurisdictional crosswalks, gap inventory, expansion plan. Vol II: discovery, retrieval, normalization, transcription, indexing, AI-assisted extraction, cross-model hostile review, verification discipline, ethical framework; techniques for recovering records a repository has stopped publishing; distinction between minutes-adoption dates and meeting dates. Vol III: partnership proposal for university-hosted research, teaching, and practitioner-training.
Cite for. Methodology and institutional case for the primary-document infrastructure behind the case-study series.
Companion. RGRC parent corpus (20088873) and Directory (20673582) are the operational sisters.
Reading paths
RQ2 · How are different kinds of electric utilities responding?
Path: Thirteen Cooperatives → Built for Someone Else → Google–CFS PPA
Extension: The Whole Record Vol II (federal utility comparison).
RQ2 · What are the costs, risks, and profits for diverse groups of utility ratepayers?
Path: Concentrated Benefit Vol I → The Whole Record Vol II → Chesterfield County
Extension: Governing the Demand Shock (Report) for the statewide fiscal framing.
RQ1 · How do diverse stakeholders perceive, contest, and participate in decision-making?
Path: Governing the Demand Shock (Report) → Henrico → Prince William → Loudoun → Louisa → Goochland
RQ3 · How do communities experience economic and environmental costs and benefits?
Path: Henrico (East-Henrico / West-End convergence) → Botetourt (Project Raspberry secrecy) → Prince William (governance capture)
Extension: The water thread — Two Systems → Aquifer Says No → Line Drawn on Highway → Where the Instrument Stops → Declared, Not Measured → No Field to Land In.
RQ3 · Worker justice and construction-trades labor
Path: Where the electricians went (stands alone as the independent quantitative measure)
Extension: Mecklenburg (the fifteen-year employment audit).
Postdoc onboarding · Which counties are documented at what depth?
Path: Commonwealth Reference (statewide inventory) → Governing the Demand Shock (Report) for the framework
Then follow the diffusion arc chronologically: Henrico → Prince William → Loudoun → Louisa → Goochland → Fredericksburg Ring → Culpeper → Botetourt → Chesterfield → Mecklenburg → Hanover → Orange.
By state
Virginia (primary coverage — statewide, analytical, and 12 county case studies)
Commonwealth Reference · Governing the Demand Shock (Report + Book) · Concentrated Benefit Vol I · Built for Someone Else · Google–CFS PPA · Thirteen Cooperatives · Two Systems · Aquifer Says No · Line Drawn on Highway · Where the Instrument Stops · Henrico · Hanover · Chesterfield · Mecklenburg · Fredericksburg Ring · Orange · Louisa · Goochland · Botetourt · Culpeper · Prince William · Loudoun Endgame · Directory · RGRC Institutional Package · RGRC Corpus
Georgia
What Georgia Is Spending, What Georgia Is Getting · Declared, Not Measured (groundwater permit triggers)
Tennessee
What Tennessee Is Building · The Whole Record: TVA Vol II
Multi-state (VA / GA / TN)
No Field to Land In (water-reporting forms) · Transmission Siting Disclosure Capture (VA/TN/TX) · Where the electricians went (US metropolitan panel, all three states in the treated set)
By actor
Investor-owned utilities
Dominion Energy: Concentrated Benefit Vol I, Built for Someone Else, Google–CFS PPA, most county case studies, water volumes (as regulated waterworks customer). Georgia Power: What Georgia Is Spending. TVA (federal corporation): The Whole Record Vol II, What Tennessee Is Building, Transmission Siting Disclosure Capture.
Electric distribution cooperatives
Thirteen Cooperatives (all 13 Virginia distribution cooperatives + Old Dominion Electric Cooperative). Mecklenburg case study touches Mecklenburg Electric Cooperative territory. The proposal names Mecklenburg Electric Cooperative in Pittsylvania County as a site.
Municipal utilities
Implicit throughout Thirteen Cooperatives (SCC regulatory distinction). Whole Record Vol II documents federal-scale but includes local-distribution structure.
Federal agencies
FERC: Concentrated Benefit Vol I (Schedule 12 cost allocation). GAO: The Whole Record Vol II (1982 report to Congress on TVA). PJM: Concentrated Benefit Vol I, Chesterfield County.
State agencies
Virginia SCC: Built for Someone Else, Thirteen Cooperatives, Chesterfield, Prince William, Google–CFS PPA. Virginia DEQ: all four water volumes. Georgia PSC: What Georgia Is Spending, Declared Not Measured.
Where the electricians went (IBEW locals across 89 metros). Mecklenburg (Microsoft labor-promise audit).
Community coalitions and advocacy
Piedmont Environmental Council (Orange, Loudoun), SELC (Directory, Prince William), Clean Virginia (Prince William governance-capture), local resident groups across every county case study, particularly Henrico (East-Henrico + West-End coalition), Louisa (AWS retreat), Prince William (democratic correction).
By decision venue
Federal (FERC / PJM / GAO / OLMS)
Concentrated Benefit Vol I (FERC Schedule 12, PJM allocation). The Whole Record Vol II (federal TVA authority, 1982 GAO). Where the electricians went (Department of Labor LM-2 filings).
Virginia SCC
Built for Someone Else (VCEA reliability exception, generation certificates). Thirteen Cooperatives (4 large-load rate schedule cases, affiliate-authority sequence). Google–CFS PPA (SCC dockets). Chesterfield (PUR-2025-00037, CERC). Prince William (PJM correction).
Georgia PSC
What Georgia Is Spending. Declared, Not Measured (though the trigger operates before Department review).
State legislative
Virginia: Va. Code § 56-234 D (2025 Acts of Assembly, chapters 499 and 598) analyzed in Thirteen Cooperatives; VCEA reliability exception analyzed in Built for Someone Else. Georgia: O.C.G.A. § 48-8-3(68.1) analyzed in What Georgia Is Spending; Ground Water Use Act analyzed in Declared, Not Measured.
State agency (water and environment)
Virginia DEQ: SJ 25 study analyzed in Aquifer Says No; annual water resources report basis for Two Systems; OWS forms analyzed in No Field to Land In. Georgia EPD: Ground Water Use Act rules analyzed in Declared Not Measured; monthly withdrawal forms in No Field to Land In. Tennessee: CN-1119 form in No Field to Land In.
Local (zoning, ordinance, permitting)
All twelve county case studies. Explicit ordinance-adoption records in Loudoun Endgame, Goochland, Louisa, Botetourt, Orange.
Courts
Prince William: Supreme Court of Virginia withdrawn filing (July 2, 2026), conflict-of-interest ruling. Botetourt: Fourth Circuit 2020 Union Hill precedent cited. Goochland: circuit court adoption-record dispute; demurrer argued October 20, 2026.
Contact for methodology or docket questions
Tommy N. Turner
Independent researcher · Advisory Board, Data Centers and the Southern Energy Buildout (Sloan / Emory / UTK)
Additional peer-reviewed manuscripts extending this corpus are currently in review at Economic Development Quarterly, Public Integrity, and Land Use Policy.