Independent civic analysisChesterfield County, Virginia

The Bermuda Hundred Review

A neutral, fully sourced examination of Google's "Project Peanut" data center campus at 2700 Bermuda Hundred Road: what is being built, what the tax deal actually says, who pays for the water and power, and a reproducible probabilistic model of what the county nets.

Sourcedtraceable to a public document or named report Judgmentreasoned estimate, stated range
No. 01

Executive summary

Google is building a data center campus called Project Peanut on roughly 307 acres at 2700 Bermuda Hundred Road in eastern Chesterfield County, beside Meadowville Technology Park. The site plan filed November 9, 2025 shows three buildings totaling about 855,846 square feet, built in phases beginning with one building on an 18 to 24 month construction schedule. It is the first of three Google campuses planned in the county (Project Skye near Moseley and Project Loch near Westchester Commons follow), together about 1,500 acres, announced August 27, 2025 as part of a $9 billion Google investment in Virginia. Sourced

The fiscal deal is public in outline, though the executed agreements are not posted online. Google pays real estate tax at the county's uniform $0.89 per $100 of assessed value, and personal property tax on its computer equipment at $0.24 per $100, a rate the Board of Supervisors locked for 30 years in a 2020 agreement, with a grant-back clause if the countywide rate ever rises. That $0.24 is the lowest among Virginia's major data center localities: Loudoun charges $4.20, Prince William $4.15, Henrico $2.60. Sourced

10-year net, median
+$34.1M
County-system net fiscal impact 2026 to 2035, real 2026 dollars. Cautious +$16.5M, optimistic +$49.4M.
20-year net, median
+$85.5M
2026 to 2045. Cautious +$31.2M, optimistic +$116.5M. The water bill lands late, which is why the long horizon matters.
Probability of net loss
≈1.3% /20yr
About 0.2% over 10 years. Worst simulated case: roughly minus $9M over 20 years.
What the ledger misses
Bills
Ratepayer-funded water capital ($426M plant) and statewide electric costs (JLARC: +$14 to $37/month by 2040) sit outside the county's general fund.

Three caveats frame those numbers. First, the upside is real but much smaller than headline framing suggests: the county's own assessment schedule depreciates computer equipment to 1 percent of cost within about five years, so the locked $0.24 rate applies to a rapidly shrinking base between refresh cycles. Second, the largest household-level risks do not run through the county's general fund at all: they run through Chesterfield's ratepayer-funded utility system, which plans a $426 million Appomattox River treatment plant for 2033, and through regional electric rates. Third, the water and power terms specific to Google are not public, so any attribution of infrastructure cost to this project is a judgment range, not a fact.

No. 02

What this means for me: a resident FAQ

Will my property taxes go down because of this?

Not automatically, and not much from this project alone. The median model outcome is roughly $4 to $7 million per year in county revenue at full buildout, against a general fund of about $1.04 billion. That is real money (roughly the operating cost of a couple of schools) but under 1 percent of the budget. Supervisors have said commercial growth helps hold residential rates down over time; Loudoun and Prince William only saw dramatic homeowner relief after accumulating dozens of data centers at tax rates 17 times higher than Chesterfield's locked rate. Sourced Judgment

Will my water bill go up?

Chesterfield Utilities is self-funded from rates and fees, and bills are already rising about 4 percent a year for system-wide needs. If the county recovers data center infrastructure costs through connection and capacity fees as its ordinance provides, the Google effect on your bill should be near zero. If a meaningful share is not recovered, our estimate is roughly $0.20 to $4.40 per household per month depending on the shortfall (the median scenario is under $1). The exact terms with Google are not public. Sourced Judgment

Will my electric bill go up?

Statewide data center growth is the main driver, not any single campus. JLARC's December 2024 study projects $14 to $37 per month in added generation and transmission costs for a typical Dominion residential customer by 2040 across all growth, and the SCC has since created a separate data center rate class to push more of those costs onto data centers. Dominion customers will also pay for the $1.47 billion Chesterfield gas peaker through a rider. Attributing a specific dollar amount to Project Peanut alone is not possible from public data. Sourced

How many jobs will it create?

Google has not said. Data centers create large temporary construction employment (plausibly a few thousand job-years here) but modest permanent staffing; a three-building campus plausibly employs 100 to 400 people directly at strong wages. The state exemption requires at least 50 jobs at 150 percent of the prevailing wage. Sourced Judgment

Was this done in secret?

The land assembly and tax agreements were approved in public votes, but the buyer's identity was hidden behind an economic development nondisclosure agreement from 2018 until August 2025. That is a common practice in site selection competitions, and it also means residents could not weigh in on the actual counterparty before key terms were fixed. Both statements are true; this analysis treats it as a structural transparency problem, not misconduct. Sourced

Can the county raise the data center tax rate later, like Henrico did?

Not for these projects. The 2020 and 2025 agreements lock $0.24 per $100 for 30 years and require the county to rebate any increase via grants. Future projects without agreements could be taxed at a higher rate. Sourced

Is the new water plant being built for Google?

The tidal Appomattox plant has been planned since at least 2019 for growth and redundancy (the January 2025 Richmond water outage reinforced the redundancy case). Data center demand accelerates and possibly enlarges the need. Attribution to Google specifically is a range from near zero to about 15 percent of Phase I capacity. Sourced Judgment

Will it be loud or polluting?

The site is industrial-adjacent (Meadowville) rather than in a subdivision, which reduces but does not eliminate noise risk; JLARC found noise problems are real in a minority of cases and recommended sound modeling. Backup diesel generators run rarely and are DEQ-permitted; regional air effects are small except in prolonged outages. Skye and Loch have documented wetland impacts under Army Corps review; Peanut has no published Section 404 action. Sourced

What happens if the AI boom busts?

The county's direct downside is limited because most revenue comes from real estate tax on buildings that exist once built, and the model's worst 2 percent of outcomes stay near breakeven (worst simulated case about minus $9 million over 20 years). The bigger exposure would be utility capacity built for demand that does not materialize, which lands on ratepayers, and that is exactly why the undisclosed water terms matter. Judgment

Who can I actually influence now?

Peanut's tax terms are locked, but building permits, site plan conditions, the Skye and Loch Army Corps permits (Skye's comment period, notice NAO-2026-00282, is open through July 28, 2026), future data center rezonings (now conditional uses countywide), utility rate cases at the SCC, and the 2035 sunset debate on the state sales tax exemption are all live processes with public comment channels. Sourced

No. 03

The project and where it stands

The site. Roughly 300 to 307 acres at 2700 Bermuda Hundred Road, bought from the county Economic Development Authority in 2020 for $18.1 million by Peanut LLC, a Google affiliate, after negotiations that began around 2018 under NDA. The parcel sits beside Meadowville Technology Park, home to the former Capital One data center (now Chirisa) and other industrial users, near I-295 and the James River. Sourced

The plan. Site plan filed November 9, 2025: three data center buildings of about 285,282 square feet each (855,846 total), 345 parking spaces, phased development starting with one building. Designer AECOM; civil engineer Bohler Engineering. Google has not disclosed megawatts, water use, cost, or jobs, citing competition. Sourced Based on the footprint and modern hyperscale densities, campus IT load plausibly lands between 150 and 400 MW at full buildout. Judgment

Status as of early July 2026. The county said in August 2025 that groundbreaking activity was underway with construction to begin by end of 2025; VPM reported in June 2026 that site plans and building permits were still under county review. Whether vertical construction has started is not clearly established in public reporting. First building operation in 2027 to 2029 is the plausible window given the 18 to 24 month build time. Sourced Judgment

Three campuses, one operator. Skye (about 871 to 887 acres at Upper Magnolia Green West near Moseley; rezoned May 2025; Dominion filings describe a 900 MW campus phased 2031 to 2033) and Loch (about 342 to 350 acres at Watkins Centre South; rezoned May 2025; a three-building campus per its federal filing) follow behind Peanut. The Army Corps public notice for Skye (NAO-2026-00282, comment period open through July 28, 2026) describes "a Google mission-critical data center campus, of no less than five data center buildings" plus three substations; that "five" refers to buildings on the Skye campus, not five separate data centers in the county. Across the three campuses the public filings describe eleven or more buildings on roughly 1,500 acres. This analysis models Peanut only, but the water, power, and rate questions are county-wide. Sourced

Timeline of key events

2018
NDA-covered negotiations begin between the county EDA and a then-anonymous operator. [Sourced]
2019
Chesterfield cuts its data center equipment rate to $0.24, billed as lowest in Virginia; the county begins planning a new water source. [Sourced]
2020
Board approves the 30-year rate-lock agreement; the EDA sells 307 acres to Peanut LLC for $18.1M. [Sourced]
Dec 2024
JLARC publishes the statewide data center study; Commonwealth Fusion Systems announces the ARC fusion plant in Chesterfield. [Sourced]
Mar to Apr 2025
FY2026 budget adopted; real estate rate cut to $0.89; the $426M Appomattox Phase I plant enters the utilities capital program. [Sourced]
May to Jun 2025
Skye and Loch rezoned; their 30-year rate locks approved unanimously. [Sourced]
Jun 30, 2025
Google signs a 200 MW fusion power purchase agreement with CFS. [Sourced]
Aug 27, 2025
$9 billion announcement; Google's identity revealed as the buyer behind all three sites. [Sourced]
Sep 2025
Zoning overhaul adopted: data centers become conditional uses countywide, effective Jan 1, 2026. [Sourced]
Nov 9, 2025
Peanut site plan filed: three buildings, 855,846 sq ft. [Sourced]
Nov 25, 2025
SCC approves a new data center rate class and the 944 MW Chesterfield gas peaker ($1.47B, rider-funded). [Sourced]
Feb 2026
SCC approves the Midlothian-to-Duval 230 kV lines ($121M) serving the Skye area. [Sourced]
Apr to Jun 2026
Army Corps review of Skye begins; community forums, petition, FOIA requests; Peanut permits still in review. [Sourced]
Jul 2026
Army Corps public notice for Skye (NAO-2026-00282): a campus of no fewer than five buildings and three substations; 17.99 acres of forested wetlands permanently impacted, 8,115 feet of streams; public comment open through July 28, 2026. [Sourced]
No. 04

Tax structure and revenue projections

Rates that apply. Real estate: $0.89 per $100 of assessed value, uniform for all taxpayers by Virginia law. Data center computer equipment: $0.24 per $100, locked 30 years (the 2020 agreement for Peanut; June 2025 agreements for Skye and Loch structured as 10 years from certificate of occupancy plus two automatic 10-year renewals tied to construction progress, a grant-back if the rate rises, a 36-month acquisition deadline, and noise and water use limits and road improvements referenced as reported). Standard computer equipment elsewhere in the county pays $3.25; machinery and tools ($1.00) does not apply to data centers. Sourced

Data center equipment tax rates, $ per $100 of assessed value (2025 to 2026)
Sources: county published rates; WVTF; VPM; InsideNoVa. Chesterfield's rate is contractually locked for 30 years; Hanover's is under a rate study that could raise it to $3.00.

The decay schedule matters more than the rate. Chesterfield's own return form (Form PP20) assesses computer equipment at 50 percent of original cost in its first tax year, then 40, 20, 10, 5, and 1 percent from year six onward. A one-time $700 million server fleet would yield about $0.8M in its best year and about $2.1M cumulatively per refresh cycle at the locked rate. Continuous reinvestment, which hyperscalers do practice, is what keeps equipment revenue alive. Sourced

Assessed value of computer equipment by age, % of original cost (county Form PP20)
Yr 1Yr 2Yr 3Yr 4Yr 5Yr 6+
Source: Chesterfield County 2025 Return of Business Tangible Personal Property.

What the county projects. Economic development staff project more than $4.2 million per year per $1 billion invested (using a $300M building, $700M equipment mix), and the EDA has cited about $71 million over 20 years per $1 billion. Sourced Simple arithmetic supports the $4.2M figure only if equipment is taxed near full cost; under the county's own declining schedule a static fleet yields far less, and the gap is closed only by sustained refresh investment. The claims are not false, but they are projections resting on refresh behavior, not contract commitments. Judgment, arithmetic shown

Our estimate for Peanut. Median cumulative county revenue of about $115M over 20 years (P10 $46M, P90 $144M), of which roughly 70 to 80 percent is real estate tax on the buildings. At maturity the median run produces about $6 to $8M per year, decaying between refresh cycles. For context, $7M is about 0.7 percent of the FY2026 general fund. Model output, judgment inputs

State treatment. Server purchases are exempt from Virginia sales and use tax through June 30, 2035 (extendable to 2040 for very large investors) under Va. Code 58.1-609.3(18), conditioned on $150M of investment and 50 jobs at 150 percent of prevailing wage. JLARC put the statewide value of this exemption at $928.6M in FY2023, Virginia's largest economic development incentive. Sourced

Revenue risks. The rate lock forecloses the Henrico path (attract at $0.40, later raise to $2.60). Real estate assessed values for data centers are contested in Northern Virginia appeals. If AI hardware cycles lengthen or workloads consolidate, refresh cadence and equipment revenue fall; AI densification could equally push investment above our ranges. Judgment

"We are the lowest in the whole state."

Supervisor Jessica Schneider, on the locked $0.24 equipment rate, VPM, June 2026
No. 05

Water: infrastructure, cost sharing, and risk

The system today. Chesterfield Utilities, an enterprise fund with no general fund tax support, has 110.5 MGD of supply capacity against 39.46 MGD average demand, from three sources: Addison-Evans on Swift Creek Reservoir (12 MGD), an Appomattox River Water Authority allocation from Lake Chesdin (66.5 MGD of the regional plant's 96), and a 32 MGD City of Richmond purchase allocation. About 122,825 active water accounts. Sourced

Water supply capacity vs. average demand, million gallons per day
Addison-Evans (Swift Creek) 12 ARWA allocation (Lake Chesdin) 66.5 City of Richmond allocation 32 Planned: Appomattox Phase I +20 (2033)
Source: Chesterfield County Utilities. Dashed line: 39.46 MGD current average demand. Allocations are contractual maximums, not firm daily yield.

The new plant. A fourth source on the tidal Appomattox River: Phase I of about 20 MGD for about $426 million, operational around 2033, inside a $1.7 billion 10-year utilities capital program; ultimate buildout about 40 MGD, with a 33-acre intake site and an 89-acre plant site. Planning predates Google (documented to 2019); the county cites growth and regional redundancy, including Richmond's January 2025 water outage. Sourced

What Google will use. Not disclosed. Del. Carrie Coyner has said county leaders plan to serve data center needs with reused non-potable wastewater; Google says only that it is working with the local authority. For scale, Google's Botetourt County campus disclosed 2 MGD growing to 8 MGD via FOIA; Peanut's filed footprint is smaller, and closed-loop or air-cooled designs use far less. Our modeling range: 0.1 to 3.0 MGD average at full campus. Sourced Judgment

Cost sharing. No executed water cost-sharing agreement is public. County ordinance requires new customers to pay connection fees including capital cost recovery scaled by meter size, with meters above 12 inches priced case by case by the utilities director. Residents' FOIA requests on this exact question were pending as of June 2026. The honest statement: the mechanism for full recovery exists; whether the negotiated terms achieve it is unverifiable today. Sourced

Risk framing. Phase I capacity attributable to Peanut: 0 to about 15 percent (0 to 3 MGD of 20 MGD). At $21.3M per MGD, attributable capital is $2M to $64M; with 50 to 100 percent recovery from Google, the unrecovered residual is $0 to about $32M in most scenarios. Our model's 20-year median for the ratepayer-side water share is about $5M (P90 about $12M). If all three campuses proceed with evaporative cooling, county-wide data center demand could claim a much larger slice of the new plant, which would make the undisclosed recovery terms the single most consequential number in this whole topic. Judgment, model output

What it could mean for a household water bill

Annualizing unrecovered capital at about 6.5 percent (30-year revenue bonds near 5 percent) across 122,825 accounts: Judgment

Estimated household impact per month, by unrecovered capital scenario
Baseline for context: average FY2026 residential utility bill is $67.89/month, already rising $2.76/month (4.2%) for system-wide needs. Chesterfield remains at least $9/month below neighboring localities. [Sourced]
No. 06

Power and other infrastructure

Grid buildout, all sourced. The Meadowville 230 kV project (SCC case PUR-2024-00179, filed October 2024) rebuilds and extends 230 kV lines and adds the new Orchard Switching Station in eastern Chesterfield, explicitly for Meadowville-area load, with phased completion through late 2028; this serves the Peanut area. In western Chesterfield, two new 230 kV lines (about 7 miles, $121M) from Midlothian substation to a new Duval substation were SCC-approved in February 2026 to serve a filed 900 MW campus (Skye) phased 2031 to 2033. The Chesterfield Energy Reliability Center, a 944 MW gas peaker at the former coal plant site near Dutch Gap, was approved November 25, 2025 at an estimated $1.47 billion recovered via a rider on all Dominion bills, briefly suspended on reconsideration and then cleared, with its air permit under appeal; target 2029.

Google's power hedge. A 200 MW power purchase agreement (June 30, 2025) with Commonwealth Fusion Systems for half the output of the planned 400 MW ARC fusion plant at James River Industrial Center, targeted for the early 2030s. First-of-kind technology; treat it as an upside option, not firm supply. Google also cites its 24/7 carbon-free energy program. Sourced

Who pays. JLARC found data centers have historically paid full cost of service, and interconnection costs are largely assigned to the customer, but system-level growth (new generation, higher energy prices, upgraded shared infrastructure) raises costs for everyone: an estimated $14 to $37 per month for a typical Dominion residential customer by 2040, in 2024 dollars, across growth scenarios. The SCC's November 2025 creation of a separate data center rate class, with minimum-payment contract requirements, is the main mitigation now in motion. Dominion says data centers "will pay their fair share." None of these figures can be attributed to Peanut alone; it is one campus inside a statewide phenomenon. Sourced

Other infrastructure. Road improvements are referenced in the Skye and Loch agreements as reported; data centers generate little traffic after construction. Fire and EMS demand is low but nonzero, since high-voltage and battery systems require specialized response capability. Our model carries $0.5M to $2.5M per year in county service costs at full campus. Sourced Judgment

No. 07

Jobs and economic activity

Construction is the dominant benefit. Three buildings over 5 to 10 years plausibly generate a few thousand construction job-years and significant local spending; JLARC found construction is where most data center economic impact concentrates. Sourced Judgment

Operations. Google has announced no numbers for any Chesterfield site. Hyperscale campuses typically staff 100 to 400 across operations, facilities, and security at strong wages; the state exemption requires at least 50 jobs at 150 percent of prevailing wage per qualifying facility. Per dollar invested, data centers are among the least job-dense land uses; per acre of industrial land they are middling; per unit of public service demand they are excellent. All three framings are true at once. Sourced Judgment

Statewide programs. Google's Virginia education and workforce commitments (AI training, free certificates, a claimed 17,000 jobs supported statewide) are real but are not Chesterfield-specific contractual commitments. Sourced

No. 08

The probabilistic model

What it estimates. The county-system net direct fiscal impact of Project Peanut only: county tax inflows (real estate, equipment personal property, one-time fees) minus county-side costs (services plus the unrecovered ratepayer share of water capacity attributable to this campus). Constant 2026 dollars, no discounting. It excludes Skye and Loch, statewide electric-rate effects, state tax expenditures, and indirect economic activity.

Method. Monte Carlo, 10,000 runs, seeded PRNG (mulberry32, seed 20260704) so the results below are computed live in your browser and reproduce exactly on every load. Horizons: 10 years (2026 to 2035) and 20 years (2026 to 2045), the latter because water capital lands around 2033.

Cumulative county-system net fiscal impact, Project Peanut
computing 10,000 runs…
10-year horizon · 2026 to 2035
Cautious · P10
Median · P50
Optimistic · P90
20-year horizon · 2026 to 2045
Cautious · P10
Median · P50
Optimistic · P90

Decomposition, 20-year medians: gross revenue about $115M; total costs about $25M, of which the water share is about $5M. All three buildings get built in about 64 percent of runs. The explicit worst case (abandonment or bust plus high water attribution and low recovery, roughly the worst 2 percent of runs) bottoms out near minus $9M over 20 years: painful but not fiscally destabilizing for a $1B general fund.

Every input, with its basis

Input distributions (triangular = min / mode / max)
InputDistributionBasis
Building 1 online2027 (35%) · 2028 (45%) · 2029 (20%)Judgment permits in review mid-2026; 18 to 24 month build
Building 2 proceeds85%, then +1 to 4 yrsJudgment hyperscaler completion record vs demand uncertainty
Building 3 proceeds80% given B2, +1 to 4 yrsJudgment
Pre-operation abandonment2%Judgment tail risk
Demand bust (refresh and expansion stop after 2032)4%Judgment explicit low-probability worst case
Real property per buildingtri $200M / $300M / $450MJudgment anchored to county $300M-per-$1B framework
Assessment ratio (real property)uniform 60% to 95%Judgment assessment practice, appeal history
Initial equipment per buildingtri $300M / $600M / $1,200MJudgment county framework midpoint, AI density upside
Ramp tranche, year 3+20% to 60% of initialJudgment
Refresh cycle / reinvestment4 to 6 yrs · 60% to 110%Judgment industry norms
Equipment assessment decay50/40/20/10/5/1% by ageSourced county Form PP20
Tax rates$0.89 RE · $0.24 equipment (locked)Sourced FY2026 rates; 2020/2025 agreements
Water demand at full campustri 0.1 / 0.8 / 3.0 MGDJudgment undisclosed; reuse plan vs Botetourt benchmark
Water capital cost$21.3M per MGDSourced $426M / 20 MGD Appomattox Phase I
Recovery from Googleuniform 50% to 100%Judgment ordinance fee mechanism exists; terms undisclosed
Annualization6.5% CRF from 2033Judgment ~5%, 30-yr municipal revenue bonds
County services$0.5M to $2.5M/yr at full campusJudgment low-service land use

Reading it honestly. The county's direct ledger on this project is very likely positive, mainly because real estate tax on three large buildings is durable. The interesting uncertainties are the size of the upside, which depends on refresh behavior under a locked ultra-low equipment rate, and the ratepayer-side exposures that this general fund ledger, by construction, does not capture. The model also cannot capture unknown terms in the unpublished agreements; if those shift infrastructure costs to the county in ways not yet public, the whole distribution shifts down.

Scenario calculator: set your own assumptions
Deterministic version of the same formulas. Buildings come online 2028 / 2030 / 2032.
10-yr net (2026-35)
real 2026 $M
20-yr net (2026-45)
real 2026 $M
Revenue at maturity
$M / year, avg 2036-45
Water bill effect
$ / household / month
Fixed here: tax rates ($0.89 and locked $0.24), county assessment decay, $21.3M per MGD water capital, 6.5% annualization, $1.5M/yr services at full campus. Nothing about this calculator changes the analysis text; it exists so you can test the assumptions yourself.
No. 09

Scenario analysis

  • Slow single building (about 10 percent of runs): one building, late equipment, minimal refresh. 20-year net roughly +$15M to +$35M. The county still comes out ahead; the real debate becomes the opportunity cost of 307 acres of prime industrial land.
  • Median path (central 50 percent): three buildings by the early 2030s, steady refresh, high fee recovery. 20-year net +$60M to +$110M; water effect on bills under $1 per month.
  • Full AI buildout (about 15 percent): equipment near the top of the ranges, fast refresh. 20-year net +$110M to +$145M, plus larger construction activity; also the heaviest water and power draw, making the undisclosed terms matter most in exactly the scenario officials hope for.
  • Bust after building 1 (about 4 percent): refresh stops after 2032; revenue decays toward the real estate floor of about $2 to $3M per year. 20-year net usually still positive (+$10M to +$40M) but far below expectations, with utility capacity risk landing on ratepayers.
  • Worst case, explicit and low probability (about 2 percent): pre-operation abandonment or early bust, high water attribution, low recovery. 20-year net between about minus $9M and +$10M. The general fund absorbs it; the lesson would be about process, not solvency.
No. 10

Risk register

Twelve tracked risks, ordered by our judgment of resident relevance
#RiskWho bears itLikelihoodSeverityNotes / mitigations
1Statewide electric cost shift ($14 to $37/mo by 2040)All Dominion customersHigh (systemic)HighSCC data center rate class (Nov 2025); minimum-payment contracts
2Equipment revenue disappoints under locked $0.24 (depreciation, slow refresh)County budgetMedium-highMediumAssessment schedule is the county's own; publish actual receipts vs projections annually
3Water terms leave capacity costs unrecoveredUtility ratepayersUnknown (terms secret)MediumPublish agreement terms; ordinance fee mechanism exists
4CERC gas peaker rider and emissionsDominion customers; nearby communitiesCertain (approved)MediumAir permit under appeal; cost is usage-dependent
5Buildout stalls (AI capex cycle)County, ratepayersLow-mediumMediumReal estate tax floor persists; phase-gate utility investment
6Wetland and stream impacts (Skye: 17.99 acres of forested wetlands permanently impacted plus 7.23 acres converted, 8,115 ft of streams, per USACE notice; Loch about 4 acres)Environment, watershed (Deep Creek, a James River tributary)Under reviewMediumUSACE Section 404 process; 43.31 wetland and 9,213 stream mitigation credits proposed; comment open to Jul 28, 2026; Peanut has no published 404 action
7Rate-lock precedent constrains future negotiatingCounty (30 years)Certain (signed)MediumApplies only to these projects; new projects face conditional use and open rates
8Transparency deficit erodes trust; litigationCounty governanceHappeningMediumFOIA responses, published agreements, community meetings with Google
9Noise and character effects near sitesNearby residentsLow-medium (Peanut is industrial-adjacent)Low-medJLARC-recommended sound modeling; conditional use rules for future projects
10Assessment appeals cut real estate revenueCounty budgetMediumLow-medConservative assessed ratios in projections
11Fusion PPA does not materialize; more grid gasRegionalMediumLow-medTreat ARC as option value; Dominion IRP fallback
12Concentration: three campuses, one operatorCounty economyStructuralLowData center share of county revenue stays small vs Loudoun's 31%
No. 11

Ratepayers, leverage, and why process matters

This section is structural, not personal. Chesterfield's officials negotiated for seven years against one of the world's most sophisticated site-selection operations, under an NDA that is standard industry practice, in competition with localities offering similar or better terms. Three structural facts deserve attention regardless of anyone's intent.

1. Asymmetry of information. Google knew its load, water needs, and investment plans; the county negotiated the rate lock in 2020 without public knowledge of the counterparty, and residents learned the terms after they were fixed. The remedy peers have adopted is publishing performance agreements and water and power terms once executed, as Henrico and Prince William largely do through budget documents.

2. Asymmetry of time. The rate lock runs 30 years; boards turn over every four. Henrico's experience (attract at $0.40, raise to $2.60 once the cluster matured) shows the value of retaining rate flexibility; Chesterfield traded that flexibility for certainty of landing the investment. Whether that trade was worth it is a policy judgment this analysis does not make; that it was a trade should be publicly acknowledged.

3. Separation of ledgers. The general fund, which looks clearly positive here, is not where the biggest risks sit. Water capital sits with ratepayers in an enterprise fund; power costs sit with every Dominion customer through SCC-regulated rates. A resident reading only the county's fiscal-impact framing would miss both. Independent review that follows all three ledgers, which is what this site attempts, is the structural fix.

No. 12

Recommendations

These are process recommendations, not a position on the project:

  1. The county should publish the executed 2020 and 2025 agreements and any water service terms with Google, redacting only legally protected specifics, and should answer the pending FOIA requests promptly.
  2. The county should publish an annual data center fiscal report: actual tax receipts by campus versus projections, utility capital attribution, and fee recovery, in the style of Loudoun and Prince William budget disclosures.
  3. Utility capacity for the campuses should be phase-gated to executed demand commitments with take-or-pay or capacity-reservation fees, so ratepayers are insulated if buildout stalls.
  4. Residents should engage the live processes: Army Corps comment periods (Skye's, under notice NAO-2026-00282, is open through July 28, 2026), conditional use hearings for future projects, SCC rate cases, and the General Assembly's 2035 sales tax exemption sunset debate.
  5. Future agreements should include job and investment milestones with clawbacks (standard performance agreement structure), sound modeling requirements per JLARC, and published water demand estimates per JLARC Recommendation 6.
No. 13

Conclusion

Project Peanut is very likely a net fiscal positive for Chesterfield County's own books: our median estimate is about $34 million over ten years and $86 million over twenty, with only about a 1 percent chance of an outright loss, because durable real estate tax on three large buildings outweighs the county's plausible unrecovered costs. It is equally true that the upside is materially smaller than the headline framing, since the locked $0.24 rate applies to equipment that the county's own schedule depreciates to 1 percent within five years; that the 30-year lock surrenders the flexibility that let Henrico raise rates sixfold once its cluster matured; and that the risks residents feel in their bills run through utility ratepayer ledgers and statewide electric rates that no county fiscal analysis captures.

The single most useful thing that could happen next is boring: publish the agreements, publish the water terms, and report actuals against projections every year.

No. 14

Author's note and independence

About the author

This analysis was put together by a resident of Botetourt County, Virginia, who prepared a similar independent analysis of the Google data center project in Botetourt. A Chesterfield County resident asked the author to complete one for this project. The author has no affiliation with Chesterfield County, Google, Dominion Energy, or any party to the project, and is not identified here by choice.

Independence disclaimer. This analysis was prepared independently. It was not commissioned, reviewed, or funded by Chesterfield County, Google, Dominion Energy, or any party to the project. Figures labeled Sourced are traceable to the public record listed below; figures labeled Judgment are the author's reasoned estimates and are presented with ranges. All errors are the author's own; corrections with sources are welcome via the feedback form.

Access note: several outlets (Richmond BizSense, Axios Richmond, VPM, WRIC) block automated access or meter articles; facts attributed to them were verified through the versions available to the compiler, including syndicated copies, and are cited by outlet and date so readers can verify directly.

No. 15

Sources

Primary and government documents

Reporting

No. 16

Corrections and feedback

Spotted an error, have a primary document, or know something the public record is missing? Submissions go directly and privately to the author (no email address is exposed; this form is processed by the site host, not a third-party mailbox).