AI Datacenter in Berkeley County, WV: The Promise They’re Selling and the Reality They’re Not

The pitch always sounds the same.

A tech company — or a developer building spec capacity for tech companies — shows up at a county commission meeting with renderings, job numbers, and the word “investment” used as many times per sentence as structurally possible. Hundreds of millions of dollars. Hundreds of jobs. A new kind of economy for a community that’s been promised economic transformation before and is still waiting.

Berkeley County, West Virginia is now in the crosshairs of AI datacenter development. The Eastern Panhandle — close to DC, served by major fiber routes, connected to PJM’s grid, and hungry for ratables — looks attractive on a site selection spreadsheet. And the pitch is already being made.

Before the county commission signs anything, before a single shovel breaks ground, Berkeley County residents deserve a clear-eyed accounting of what these facilities actually deliver, what they actually cost, and why the communities that get them rarely end up ahead.

The Project: What We Know So Far

Based on public reporting — including coverage by Data Center Dynamics — and submissions to state agencies, here is what has been disclosed about the specific project under development:

The Bedington Campus, as it is being called by its developer, is planned for approximately 548 acres in the Falling Waters District of Berkeley County — in the Eastern Panhandle, not far from the Maryland border. The site sits near the Falling Waters and Bedington communities, roughly southwest of Martinsburg, with proximity to Interstate 81 and existing power infrastructure cited as key siting factors.

The developer is Penzance Management, which has already purchased the land through a commercial real estate transaction — this was not a grant, donation, or government-conveyed transfer. The deed of sale is a matter of public record through the Berkeley County Clerk’s office, as all property transfers in West Virginia are recorded with the county assessor and clerk. Penzance is now in the process of preparing the site for infrastructure development. Environmental and engineering planning has been submitted to state agencies as part of the approval and certification process, though comprehensive project details have not been made fully public at this stage.

At full buildout, the campus is projected to encompass approximately 1.9 million square feet of facilities dedicated to data center operations and related infrastructure — a footprint comparable to several large shopping malls combined. The facility is designed to deliver up to 600 megawatts of IT power capacity, making it one of the largest AI infrastructure projects in the Mid-Atlantic region if realized.

Officials have indicated that the project may use reclaimed water for cooling as a mitigation measure for environmental impact, though detailed water sourcing and infrastructure plans have not been fully disclosed. Community awareness is growing: residents in the area have noted the general vicinity around Scrabble Road and Beddington Road as part of the local discussion, and public meetings and county hearings are being scheduled to address community questions and concerns.

These are the facts on the table. What follows is an honest accounting of what they mean.

What They Promise

The datacenter industry runs on a standard playbook when approaching local governments. The promises fall into three buckets: jobs, tax revenue, and economic spillover.

Jobs: The number cited is almost always construction jobs — temporary by definition. Permanent operational headcount for a modern AI-scale datacenter typically runs between 50 and 200 people for a facility that might occupy several hundred thousand square feet. Those positions tend to require specialized skills in electrical systems, networking, and facilities management. They are real jobs. They are not the community employment transformation that gets implied when a company talks about “hundreds of jobs.”

Tax revenue: Datacenter operators are aggressive — and often successful — at negotiating property tax abatements, sales tax exemptions on equipment, and other incentive packages that significantly reduce the fiscal contribution that would otherwise accrue to local government. West Virginia has competed hard for datacenter investment in part by offering favorable tax treatment. The question of whether the net fiscal benefit after incentives covers the infrastructure burden is one that economic development agencies rarely model honestly.

Economic spillover: The “multiplier effect” — the idea that datacenter employees will shop locally, eat locally, and generate downstream economic activity — is real but modest. A facility with 100 permanent employees in a county of 120,000 people does not transform the local economy. It adds some payroll and a handful of lunch spots get a marginal boost.

That’s the honest version of the benefits. It is less exciting than the pitch. It is closer to the truth.

What They Don’t Lead With

The costs are real, they are borne primarily by the community rather than the company, and they are rarely front-and-center in the economic development presentation.

Water

AI training and inference at scale generates enormous amounts of heat. Modern hyperscale facilities use water-cooled systems that consume millions of gallons annually. A large facility can require anywhere from 1 to 5 million gallons of water per day depending on design and climate.

The Eastern Panhandle is not drought-prone by historical standards, but the Potomac watershed faces growing pressure from regional population growth, agricultural demand, and the compounding effects of climate variability. A facility drawing millions of gallons per day from local water systems — or from groundwater — creates real competition with residential and agricultural users. Water rights and water infrastructure capacity are questions that need hard answers before a permit is issued, not after.

When datacenters locate in arid regions without adequate water planning, the consequences for surrounding communities have been documented and severe. The same conversation needs to happen here, with Berkeley County’s water authority at the table.

Officials connected with the Bedington Campus have floated the use of reclaimed water for cooling — treated wastewater reused for industrial cooling purposes rather than drawing directly from potable supplies. This is a meaningful mitigation strategy if properly implemented, and it deserves credit as an acknowledgment that water demand is a real concern. But “might use reclaimed water” is not a binding commitment, reclaimed water infrastructure requires its own investment and planning, and the Falling Waters area’s current reclaimed water capacity and proximity to the proposed site have not been publicly detailed. A promise of reclaimed water use without a specific sourcing plan, infrastructure commitment, and enforceable permit condition is not a water solution — it is a talking point.

The Aquifer Beneath Your Feet — and Why It Makes This Worse

Before any discussion of water consumption can be had honestly, Berkeley County residents need to understand what is actually under the ground they’re standing on — because it changes the stakes of this conversation considerably.

The Eastern Panhandle sits atop one of the most significant carbonate rock aquifer systems in the Mid-Atlantic region. The subsurface geology here is dominated by ancient limestone and dolomite formations — the Tomstown Dolomite, the Beekmantown Group carbonates, and the Conococheague Limestone among them — deposited during the Cambrian and Ordovician periods roughly 450 to 500 million years ago. Hundreds of millions of years of slightly acidic groundwater moving through fractures and bedding planes have dissolved these carbonate rocks into an intricate underground network of caves, conduits, enlarged fractures, and interconnected voids.

This is karst terrain. And Berkeley County is deeply, extensively underlain by it.

The result of this geology is a groundwater resource of unusual productivity. Wells drilled into Berkeley County’s carbonate aquifers have historically tested at yields that dwarf what most of the country considers a productive well — individual municipal production wells in the county have recorded yields exceeding 1,000 gallons per minute. The Opequon Creek watershed, the Warm Springs Ridge corridor, and the lowland valleys along the Eastern Panhandle’s limestone belts overlie aquifer zones that have supported agriculture, industry, and a rapidly growing population for generations. The water stored in fractures, conduits, and pore spaces across these formations represents a resource accumulated over geologic time — a buried resource of staggering scale that has historically been treated as essentially inexhaustible.

A large AI datacenter drawing 1 to 5 million gallons per day from this system would stress that resource in ways that current residents have never experienced. But the water volume is only part of the risk. The more serious problem is what karst terrain does when you start changing the rules.

Karst Geology and the Risk of Ground Collapse

Karst landscapes are inherently unstable in ways that standard geotechnical assessments frequently miss — or underweight.

The same dissolution processes that make Berkeley County’s aquifers so productive have also created a subsurface riddled with cavities of every size: from hairline fractures to rooms large enough to walk through. Many of those cavities exist at shallow depths — tens of feet below the surface — held in precarious equilibrium by the natural groundwater pressure that fills them and the thin rock or soil bridge that spans them. In a stable natural system, this equilibrium can persist for centuries. Disturb it, and the results can be sudden and severe.

Sinkholes in karst terrain are most commonly triggered by:

Rapid groundwater table drawdown. When large volumes of groundwater are extracted quickly, the buoyant support that saturated conditions provide to shallow cave ceilings and soil arches above voids is removed. The unsupported roof material loses strength, and collapse follows — sometimes days after pumping begins, sometimes years later. The Eastern Panhandle’s West Virginia Geological and Economic Survey records document numerous sinkhole and subsidence events associated with drought conditions and well drawdown in the carbonate rock belt. Industrial-scale extraction would be a drought of local, human-caused, unrelenting duration.

Concentrated surface water discharge. Cooling towers don’t just consume water — they discharge warm, conditioned water that has to go somewhere. When that discharge volume is directed into the ground in concentrated areas, or when stormwater from large impervious surfaces is channeled to specific infiltration points, it accelerates dissolution of carbonate rock and can rapidly destabilize subsurface voids that were previously stable. The karst system doesn’t distinguish between natural recharge and industrial discharge. It responds to chemistry and hydraulics.

Heavy static loading. A hyperscale datacenter building — its concrete foundations, raised floor systems, server racks, power infrastructure, and cooling equipment — represents a static load of tens of thousands of tons placed on what is, in karst terrain, a structurally heterogeneous subsurface. Ground beneath a large building in karst country may include solid rock inches away from a void large enough to swallow a house. Standard geotechnical borings, spaced at even 25-foot intervals, can completely miss significant cavities. When the load exceeds the span capacity of a shallow cavity roof, the collapse is not gradual. It is sudden.

Vibration from mechanical equipment. Large cooling systems, diesel generators, and HVAC infrastructure generate persistent low-frequency vibration. In karst terrain, sustained vibration can loosen the unconsolidated fill material that bridges shallow cavities — a process called raveling — and gradually work that material into underlying voids until the surface structure above it has nothing left to stand on.

West Virginia’s karst collapse history is not theoretical. The state has documented dozens of significant sinkhole events in Eastern Panhandle counties, affecting roads, utility corridors, and structures. Counties across the Appalachian carbonate belt — from Pennsylvania through Virginia and into Tennessee — have well-documented records of industrial and residential development triggering sinkhole events that were not predicted by conventional site assessments.

The liability question is one no one in the economic development presentation will raise: if a datacenter’s groundwater extraction or surface loading contributes to sinkhole formation beneath a neighboring road, utility line, or home, who is responsible? West Virginia’s framework for addressing karst-related liability in industrial development contexts is underdeveloped relative to the geological risk present in this region.

Any serious consideration of a datacenter in Berkeley County should require a full karst hazard assessment — not a standard Phase I environmental review, not a conventional geotechnical report, but a dedicated karst investigation including geophysical surveys, sinkhole inventory mapping, groundwater tracer studies, and probabilistic cavity location analysis. This is the industry standard for karst-prone development in states that take the risk seriously. West Virginia should demand it here.

The aquifers beneath Berkeley County took millions of years to form. The landscape above them reflects that geology in ways that development plans often don’t account for. Before the county invites an industrial water consumer of this scale to set up operations over one of the most productive — and most geologically complex — aquifer systems in the region, it deserves to know exactly what it’s risking.

What’s in the Water After It Cools the GPUs

The cooling water conversation almost never gets to the part that matters most to downstream users: what is in that water after it has done its job?

The answer is: a lot of things that were not in it before.

Modern GPU clusters — the kind that run AI training and inference at scale — generate extreme localized heat. NVIDIA’s H100 and H200 cards, the workhorses of current AI infrastructure, can dissipate over 700 watts each under sustained load. Racks dense with these chips can reach tens of kilowatts. The industry has increasingly moved toward direct liquid cooling (cold plates bonded directly to GPU packages) and immersion cooling (servers submerged in dielectric fluid) to handle these thermal densities. In all of these systems, the fluid doing the cooling ultimately picks up chemical baggage before it is discharged, reused, or treated.

Cooling tower blowdown is the most significant discharge pathway. As water evaporates in a cooling tower, dissolved minerals concentrate in the remaining water. When that concentration reaches a threshold — typically three to five times the starting concentration — the concentrated water is “blown down” (discharged) and replaced with fresh makeup water. That blowdown stream carries:

Scale inhibitors and corrosion inhibitors: Phosphonate compounds, azole derivatives (benzotriazole and tolyltriazole are the two most common), zinc salts, and molybdate compounds are routinely added to cooling water to protect heat exchanger surfaces. Benzotriazole and tolyltriazole have been classified as contaminants of emerging concern by the U.S. EPA; they are highly water-soluble, resistant to biodegradation, and have been detected in drinking water systems downstream of industrial cooling discharge. They are not currently regulated under federal drinking water standards, which means most utilities do not test for them.
Biocides: Cooling water is a near-perfect growth medium for microbial communities — warm, nutrient-containing, and circulating. Facilities dose their systems aggressively with biocides to suppress Legionella, algae, and biofilm. Common agents include chlorine dioxide, bromine compounds, quaternary ammonium compounds (quats), glutaraldehyde, and isothiazolinones (CMIT and MIT). These are acutely toxic to aquatic organisms at low concentrations. They do not disappear when the water is discharged — they persist in the discharge stream at levels that depend on dosing rate, system residence time, and treatment technology.
Heavy metals: Corrosion of copper heat exchangers, aluminum server chassis, and galvanized components introduces copper, zinc, lead, and nickel into the cooling circuit. In a closed loop, these metals concentrate over time and are present in blowdown at elevated levels. Copper is acutely toxic to aquatic invertebrates at concentrations measurable in parts per billion.
Concentrated dissolved solids (TDS): The blowdown stream carries hardness minerals, sulfates, chlorides, and whatever else was dissolved in the makeup water — now concentrated several-fold. High TDS discharge into surface waters disrupts the ionic balance of aquatic ecosystems.
Thermal loading: Even “treated” cooling tower discharge is substantially warmer than the receiving water. Elevated water temperature reduces dissolved oxygen, stresses cold-water fish species, accelerates the metabolic rate of pathogens, and can trigger algal blooms in nutrient-rich waters. A facility discharging millions of gallons per day of warm blowdown into a watershed already under stress creates measurable thermal gradients in receiving streams.

For direct liquid cooling and immersion cooling systems specifically, additional compounds enter the waste stream:

Glycol antifreeze compounds: Ethylene glycol and propylene glycol are widely used as heat transfer fluids in cold-plate systems. Their breakdown products — glycolate, formate, and oxalate — are oxygen-depleting in waterways. Ethylene glycol is toxic to mammals (it is the active component of automotive antifreeze) and its sweetness makes it an attractive poison to wildlife if discharge is uncontrolled.
Dielectric fluids: Immersion cooling systems use mineral oil, synthetic esters, or fluorinated fluids (including some PFAS-containing formulations) as coolants. Fluorinated dielectric fluids from earlier generations of systems — some 3M Novec products — contained perfluoroalkyl substances. Although major data center operators have largely transitioned away from the most persistent PFAS-containing fluids, the industry’s use of complex fluorinated chemistry in cooling circuits is not fully disclosed and warrants scrutiny in permitting.
Metal particulates: GPU cold plates are bonded to silicon die packages using thermal interface materials. Copper, indium, and solder alloy particulates can enter the coolant stream through system wear and maintenance operations.

Legionella and aerosol risk deserves separate mention. Cooling towers release water vapor as drift — tiny droplets that escape into the ambient air even with drift eliminators in place. That aerosol carries whatever is in the cooling water, including live Legionella bacteria if the biocide program is inadequate. Legionella pneumophila, the causative agent of Legionnaires’ disease, thrives at the temperatures common in cooling tower basins. Outbreaks associated with cooling tower drift have been documented at industrial facilities across the United States. The risk increases with proximity of residential population to the facility — and in the Falling Waters area, residential development is present.

The karst connection makes all of this worse. In a conventional soil environment, contaminated surface water moving toward groundwater passes through soil horizons that provide some filtering and attenuation. In a karst aquifer — the geology directly underlying the Falling Waters District — there is no such buffer. Stormwater runoff carrying cooling system blowdown, biocide residuals, and heavy metals can enter surface sinkholes or losing stream reaches and reach active groundwater conduits within hours. Tracer studies in Appalachian carbonate aquifers have documented groundwater velocities exceeding one mile per day in conduit-flow zones. A spill or discharge event that would be an isolated surface water impact in most geologic settings becomes a direct aquifer contamination event in karst terrain.

The proposed use of reclaimed water for cooling — mentioned by project officials — would reduce freshwater consumption but does not eliminate the discharge chemistry problem. Reclaimed water typically contains higher initial concentrations of nutrients, pharmaceuticals, and other residuals than surface water; when concentrated through evaporation cycles and treated with biocides, the resulting blowdown stream can be more challenging to manage than blowdown from a freshwater-sourced system.

None of this is disclosed in the economic development pitch. The water chemistry of a hyperscale datacenter’s discharge stream is an industrial wastewater management question that belongs in the permitting process — specifically under a West Virginia National Pollutant Discharge Elimination System (NPDES) permit from the Department of Environmental Protection. The public should know what discharge limits will be imposed, what monitoring will be required, where the discharge will go, and what the pathway to groundwater is given the karst geology beneath the site.

Power

Welcome to the AI Data Center — Running 100,000 AI Models. Burning electricity like it's going out of style.

This is the big one, and it’s the one the industry is most reluctant to discuss honestly.

AI workloads are extraordinarily power-intensive. A large AI training cluster can draw hundreds of megawatts. Even a moderately-sized inference facility — the kind that runs AI queries rather than trains models — draws tens of megawatts continuously. These are not small power draws. They are utility-scale loads.

In the case of the Bedington Campus specifically, the projected IT power capacity is 600 megawatts — a figure that, to put it in context, is comparable to the output of a mid-sized natural gas peaker plant, or enough electricity to power roughly 450,000 average American homes. That is not a rounding error in the regional grid. It is a transformative new load.

The PJM Interconnection — the grid serving most of the Eastern Panhandle — is already under significant load growth pressure from datacenter concentration in Northern Virginia. Dominion Energy and other utilities have been candid: the rate of datacenter load growth is outpacing their ability to build generation and transmission capacity. The result is grid congestion, capacity constraints, and upward pressure on electricity rates across the region.

When a new large industrial load connects to a distribution system, the infrastructure upgrades required — new substations, transmission line upgrades, transformer replacements — are costs that get socialized across the ratepayer base. The company gets power at negotiated industrial rates. The neighboring households pay higher bills to subsidize the grid investment that made the facility possible.

This is not speculation. It is the documented pattern in Virginia, where a decade of datacenter concentration in Loudoun County has contributed to rate pressures that affect every residential and small commercial customer on the Dominion system.

What makes this worse is that Berkeley County already has a power reliability problem that no datacenter pitch will acknowledge. Residents throughout the Eastern Panhandle are well acquainted with what locals call “power blips” — brief voltage interruptions that don’t last long enough to trip most circuit breakers but are long enough to reset clocks, reboot computers, crash unsaved work, and damage sensitive electronics. These are not rare anomalies. When they occur, they can hit multiple times in rapid succession — three or more blips within a single day is not unusual for affected areas of the county.

For a home or small business, a power blip is an irritating fact of life. For a facility drawing 50 to 200 megawatts continuously — one whose entire business model depends on uninterrupted compute — a momentary voltage event is a catastrophe. Servers crash mid-operation. Cooling systems restart. Transactions fail. Storage arrays enter recovery modes. The datacenter industry addresses this reality with massive battery backup systems, diesel generators, and redundant utility feeds. Those systems are expensive to build, generate additional emissions on-site, and represent infrastructure burdens the company will negotiate into its rate agreements — which means the community’s pre-existing grid instability becomes yet another cost that ultimately flows back toward ratepayers.

The more consequential question is what happens to power quality and reliability when you attach a massive new industrial load to a distribution system that already struggles with stability. Grid operators work to maintain reliability, but adding hundreds of megawatts of demand to infrastructure that already delivers multiple daily voltage events to residential customers is not a path toward improvement for those customers sharing the same lines.

Berkeley County residents should ask: who pays for the grid upgrades? What are the projected impacts on residential electric rates? What generation resources will serve this load, and what are the emissions implications? And specifically: what is the current reliability profile of the substations that would serve this facility, and what is the utility’s concrete plan to address the power blip frequency that existing residential and commercial customers already experience?

Property Values and Quality of Life

Modern hyperscale datacenters are large, loud, and visually industrial. They generate constant mechanical noise from cooling systems — a low, persistent hum that carries further than people expect and doesn’t stop at night. They require significant truck traffic for equipment deliveries and maintenance operations. They are lit at night for security. They occupy large footprints that preclude other uses.

For residents in the surrounding area, the quality of life impacts are real. Neighboring property values in residential areas adjacent to industrial datacenter campuses have shown mixed results — some studies show neutral effects, others show depression of values particularly for properties within a half-mile of cooling infrastructure.

In the case of the Bedington Campus, the 548-acre site sits in the Falling Waters District — an area with established residential communities in the vicinity. Residents around Scrabble Road and Beddington Road have already begun raising questions through local channels, and county hearings are being scheduled. This is not a remote desert site buffered by miles of unpopulated land. The neighbors are already here, and they will live with whatever gets built.

None of this is in the economic development presentation.

Community Capture Without Community Benefit

Here is the deeper structural problem, and it is the one that should give Berkeley County pause most of all.

Datacenters are capital-intensive, highly automated industrial facilities. They are not woven into the community fabric the way a manufacturing plant with 500 line workers is. They don’t require local supply chains. They don’t generate meaningful local vendor spending. The executives and engineers who run them often live elsewhere and commute or work remotely. The physical infrastructure — servers, networking equipment, power systems — is procured through national and global supply chains.

The community provides the land, the water access, the grid connection, the permitting, the road infrastructure, and frequently the tax incentives. The company takes the product of AI compute and sells it to customers who are not in Berkeley County, to users who are not in Berkeley County, generating profits that flow to shareholders and executive compensation that is not in Berkeley County.

This is the extractive model. The community absorbs the costs and the externalities. The company captures the value.

This isn’t unique to West Virginia. It is the standard datacenter deal structure. And the communities that have accepted it — and then watched the job numbers fail to materialize, watched their power bills rise, watched the company return to renegotiate the tax abatement at renewal — have learned this lesson the hard way.

Where These Things Actually Belong

There is a reason the largest hyperscale datacenter clusters in the world are located in eastern Oregon, the high desert of Utah, the outskirts of Phoenix, and the plains of Iowa: land is cheap, water (for some cooling approaches) or cold air is available, and — critically — the population density is low enough that the impacts are distributed across far fewer people.

The desert siting argument is not simply aesthetic. It is practical.

In low-population desert environments:

Power grid impacts are smaller because the baseline load is lower and transmission corridors can be built with less conflict
Water sourcing can use innovative approaches — closed-loop cooling, aquifer recharge programs, brackish water — that are more viable when you’re not competing with dense suburban water systems
Noise and light impacts affect fewer residents
Land acquisition is straightforward and doesn’t consume agricultural or residential acreage
The fiscal math sometimes actually works because infrastructure costs are lower and the facility represents a larger share of the local tax base

Microsoft, Google, Amazon, and Meta have all built major capacity in the desert Southwest and the rural Northwest for exactly these reasons. When they come to suburban or exurban markets like the Eastern Panhandle, they are coming because they need low-latency proximity to the DC/NoVA market — not because Berkeley County is the ideal host.

The community is being asked to absorb the costs of proximity so that AI companies can shave 5 milliseconds off their network latency to serve Northern Virginia clients. That is an honest framing of the transaction.

Penzance’s Track Record: What Residents Should Know

Berkeley County residents have every right to investigate the developer’s record before welcoming a 548-acre industrial campus into their community. Here is what is publicly documented and what warrants further scrutiny.

Who Is Penzance?

Penzance (also operating as Penzance Management or Penzance Companies) is a Washington, D.C.-based real estate investment and development firm that has historically focused on commercial office and mixed-use properties in the DC metro area — Northern Virginia, DC proper, and the Maryland suburbs. The West Virginia data center campus represents a significant pivot toward hyperscale industrial development.

The company is privately held, which limits the public disclosure required of publicly traded real estate investment trusts (REITs) or publicly listed developers. Privately held developers are not subject to the same SEC reporting requirements, meaning community members cannot easily review financial filings, litigation disclosures, or material risk statements that would be available for a publicly traded company.

Legal Disputes and Community Opposition

The Bedington Campus project is early-stage, and comprehensive litigation records require direct searches of court dockets — specifically the Berkeley County Circuit Court, the Northern District of West Virginia federal court (via PACER), and the West Virginia Supreme Court of Appeals docket. As of this writing, formal legal challenges may still be in their formative stages, as the project has not yet reached the permitting stages that typically trigger litigation.

What is documented:

Community opposition has already formed. Residents in the Falling Waters and Beddington Road corridor have raised concerns through local channels. County hearings are being scheduled. The level of organized opposition in the early pre-permit stage is notable — it suggests that when formal proceedings begin, legal challenges are likely.
West Virginia’s karst terrain creates grounds for legal challenge. Development in karst terrain carries documented environmental and geotechnical risks. If Penzance proceeds without a comprehensive karst hazard assessment, that creates a concrete basis for challenge under West Virginia’s Environmental Protection Act and potentially under the federal Clean Water Act, which requires consideration of groundwater connectivity in karst settings.
Water rights disputes in the Eastern Panhandle are not hypothetical. The Morgan County and Jefferson County areas of the Eastern Panhandle have seen prior litigation over groundwater rights and industrial water draw. A facility drawing millions of gallons per day in a karst aquifer setting could trigger challenges from the Potomac Riverkeeper Network, the West Virginia Rivers Coalition, or private landowners with documented well interference.
Community members seeking legal information should consult the Berkeley County Circuit Court Clerk (public docket access), the West Virginia Department of Environmental Protection (permit applications and public comments), and the EPA’s ECHO database (echo.epa.gov) for any enforcement history associated with Penzance-affiliated entities or the specific property address.

EPA Violations and Environmental Compliance

A search of the EPA Enforcement and Compliance History Online (ECHO) database (available at echo.epa.gov) for Penzance and affiliated entities should be a standard step for any due diligence process around this development. This database provides publicly accessible records of:

Clean Air Act violations
Clean Water Act (NPDES) permit violations
Resource Conservation and Recovery Act (RCRA) violations
Superfund (CERCLA) site associations

What community members and journalists should specifically investigate:

Prior Penzance development sites — any commercial or mixed-use properties in Penzance’s DC-area portfolio that are associated with stormwater permit violations, unauthorized discharge, or sediment and erosion control failures during construction. Construction-phase NPDES violations are the most common compliance issues for large development projects.
The Bedington Campus site itself — the Berkeley County parcel may have prior use history (agricultural, commercial, or otherwise) that could carry existing environmental liabilities. A Phase II Environmental Site Assessment should be in the public record or obtainable through FOIA.
West Virginia DEP records — the West Virginia DEP PERMTS database and its enforcement records are the state-level equivalent of EPA ECHO for projects regulated under West Virginia law.
Stormwater and grading permits — large site preparation activities (land clearing, grading) require West Virginia Construction Stormwater permits. Any violations of those permits during site prep would be recorded with WV DEP and would be a matter of public record.

An important note on data currency: Because this project is actively in its early planning and permitting phase, environmental enforcement records will be sparse or nonexistent for this specific facility as of early 2026. The absence of current violations is not a clean bill of environmental health — it reflects that regulatory scrutiny intensifies as construction and operations begin, not before. The public comment period on any NPDES or air permit application is the critical window for community input on discharge limits and environmental conditions.

If you are a Berkeley County resident and want to track this project:

West Virginia DEP permit search: dep.wv.gov
EPA ECHO enforcement database: echo.epa.gov
Berkeley County Commission meeting minutes: berkeleywv.org
PACER federal court docket search: pacer.gov
Berkeley County Circuit Court Clerk: 100 West King Street, Martinsburg, WV 25401

The Honest Question to Ask

Before Berkeley County’s elected officials commit to any datacenter deal — before any tax incentive is authorized, before any water or power infrastructure commitment is made — they should require answers to these questions:

What is the permanent operational headcount, and what percentage of those jobs will be filled by current Berkeley County residents versus imported labor?
What are the specific water draw projections, and how does that compare to current system capacity and drought-year reserves?
Who pays for grid upgrades, and what is the projected impact on residential electricity rates over the next ten years?
What is the net fiscal impact after all incentives, abatements, and infrastructure costs — not the gross investment number, the net?
What are the noise, light, and traffic impacts on adjacent residential areas, and what mitigation is legally binding in the permit?
What is the current power reliability profile of the substations and distribution circuits serving the proposed site, and what is the utility’s specific plan to address the power blip frequency that Berkeley County customers already experience — before adding a massive new load to the system?
What are the clawback provisions if the facility fails to deliver on job commitments?
What is the company’s track record in comparable communities — what did they actually deliver versus what they promised?
Has a full karst hazard assessment been completed, including geophysical surveys, sinkhole inventory mapping, and groundwater tracer studies — and what are the findings?
What cooling water chemistry will be used, and what are the projected discharge volumes, temperatures, and concentrations of biocides, azole compounds, heavy metals, and glycol compounds in the blowdown stream? What NPDES permit limits will apply, and where does the discharge go relative to karst recharge zones and public water supply intakes?

If the answers to those questions are satisfactory, then a datacenter might be the right deal for Berkeley County. But the burden of proof should be on the company making the pitch, not on the residents who will live with the consequences.

Conclusion

I want to be clear about where I stand: I think AI datacenters coming to populated communities like Berkeley County do more harm than good, and I think the industry’s pitch routinely obscures that reality behind inflated job numbers and investment figures that don’t survive scrutiny.

The Bedington Campus — 548 acres, 1.9 million square feet, 600 megawatts — is not a modest pilot program. If it reaches the scale being planned, it will be one of the largest AI infrastructure deployments in the region. The impacts on Berkeley County’s water, power grid, geology, and residential communities will not be modest either. Penzance Management purchased the land. Environmental planning is in progress. This is real, and it is moving forward on a developer’s timeline.

The jobs don’t materialize at the scale promised. The tax revenue gets negotiated away. The power costs get socialized. The water gets consumed. The noise becomes permanent. And the profits leave.

These are industrial facilities. They belong in industrial settings — remote, low-population areas where the land is cheap, the impacts are diffuse, and the communities entering the deal are doing so with clear eyes rather than economic desperation.

Berkeley County is a growing community with real assets: its proximity to the federal corridor, its existing workforce, its transportation infrastructure. Those assets are worth more than datacenter promises. The county should be selective about what it invites in and who it asks to absorb the costs.

When a company wants to put something in your backyard, the most important question is always: why your backyard?

Make them answer it.

This post represents the author’s opinion based on publicly available data about the datacenter industry’s community impacts. Project-specific details about the Bedington Campus — including the 548-acre site, 1.9 million square foot buildout, 600 MW power capacity, and Penzance Management’s involvement — are drawn from public reporting and submissions to West Virginia state agencies as of March 2026. The land transaction was a commercial real estate purchase; deed records are publicly available through the Berkeley County Clerk’s office at the Berkeley County Courthouse (100 West King Street, Martinsburg, WV 25401) and through the West Virginia State Auditor’s property records system. Details remain subject to change as the approval process proceeds. Readers are encouraged to consult their local elected officials, attend public hearings, and demand transparency in any economic development agreements made on their behalf.

References and Further Reading

Penzance Targets 600MW Data Center Campus in West Virginia — Data Center Dynamics
EPA Enforcement and Compliance History Online (ECHO) — search for Penzance or the Falling Waters site address
West Virginia DEP Permit and Compliance Records
Berkeley County Commission — public meeting minutes and agendas
Potomac Riverkeeper Network — regional water quality advocacy
West Virginia Rivers Coalition — statewide water resource advocacy
PACER Federal Court Docket Search — Northern District of West Virginia
West Virginia Geological and Economic Survey — Karst Resources