Impact Report · Plymouth Township, PA
850 MW Data Center
The Real Cost
A proposed 850 megawatt data center at 900 Conshohocken Road would be one of the largest industrial facilities ever built in Montgomery County - here is what that means for you, your family, and your community.
900 Conshohocken Rd · Plymouth Township · Montgomery County, PA · PECO/PJM Grid · Schuylkill River 0.4 mi
850 MW
Continuous power draw
2.1M
Tons of CO₂ per year
7.1M
Gallons of water daily
2M
People downstream on Schuylkill
01
Turbine Methodology - What 850 MW Actually Requires
The equipment choice determines every impact figure in this report
* Methodology note - estimates and assumptions
All figures in this report are independent estimates based on publicly available engineering specifications, industry benchmarks, and regulatory data. Specific assumptions: turbine output based on GE Vernova LM6000PC/PG and LM2500XPRESS published datasheets at ISO conditions (59°F, sea level, natural gas fuel); actual output will vary with ambient temperature, inlet losses, and elevation. Generator counts assume N+1 redundancy per Uptime Institute Tier III/IV standards using 2-4 MW diesel gensets per Cummins, Caterpillar, and Generac data center specifications. Emissions calculated using EPA eGrid 2023 PA-specific factor (0.29 metric tons CO2/MWh) and GE Vernova published NOx figures (15 ppm DLE mode). Water consumption assumes PUE of 1.3 and wet-cooled tower rate of 1.8 gal/kWh per Lawrence Berkeley National Laboratory data center water benchmarks. Health and noise impacts cited from WHO Environmental Noise Guidelines, EPA standards, and peer-reviewed literature referenced in sources. Tax figures drawn from PA Department of Revenue program guidelines, Spotlight PA reporting (Feb 2026), and Data Center Dynamics coverage of this specific application. These are estimates. The developer has not disclosed actual specifications. Independent environmental review would be required to produce site-specific figures.
Primary generation needed
850 MW
Continuous, primary load
Most likely turbine
GE LM6000
46–52 MW per unit
Units required (LM6000)
17–20
Primary + N+1 spare
Diesel backup gensets
215–430
Not 40 as claimed
GE Vernova LM6000
Most likely primary turbine - the dominant aeroderivative in this power class
Output: 46–52 MW per unit (LM6000PC/PG variants). 5-minute cold start. 99%+ operational reliability. Over 1,300 units shipped globally with 40M+ operating hours. Derived from GE's CF6-80C2 jet engine.
Why this model fits this site: Compact footprint (4.9 m long) fits inside existing steel mill bays, matching the proposal to place turbines "within and between buildings." Currently the #1 aeroderivative turbine for data center onsite generation in North America. GE Vernova has an explicit data center turbine program. Crusoe Energy, the leading data-center-native power operator, uses GE Vernova exclusively.
Units needed: At 46–52 MW each, 850 MW requires 17–19 units for primary generation plus 1–2 spare units for N+1 redundancy = 18–21 turbines total.
Most probable choice - confirmed in comparable deployments
Why this model fits this site: Compact footprint (4.9 m long) fits inside existing steel mill bays, matching the proposal to place turbines "within and between buildings." Currently the #1 aeroderivative turbine for data center onsite generation in North America. GE Vernova has an explicit data center turbine program. Crusoe Energy, the leading data-center-native power operator, uses GE Vernova exclusively.
Units needed: At 46–52 MW each, 850 MW requires 17–19 units for primary generation plus 1–2 spare units for N+1 redundancy = 18–21 turbines total.
GE LM2500XPRESS
Alternative - 35 MW units, the data center industry's emerging standard
Output: 35 MW per unit. 14-day installation from flat-packed modules (20 workers). Selective Catalytic Reduction (SCR) emissions control. 5-minute fast start.
Why it's relevant: This is the exact model Crusoe Energy ordered - 29 units for ~1 GW across multiple data center sites (June 2025). Designed for rapid behind-the-meter deployment precisely at sites like this one where grid connection is delayed or insufficient.
Units needed: At 35 MW each, 850 MW requires 25–27 units for primary generation plus 2–3 spares = 27–30 turbines total.
Lead time problem: New GE Vernova units have a 3–5 year backlog. the applicant would likely need refurbished units or the LM6000VELOX fast-delivery variant.
Confirmed in real data center deployments - 2024–2025
Why it's relevant: This is the exact model Crusoe Energy ordered - 29 units for ~1 GW across multiple data center sites (June 2025). Designed for rapid behind-the-meter deployment precisely at sites like this one where grid connection is delayed or insufficient.
Units needed: At 35 MW each, 850 MW requires 25–27 units for primary generation plus 2–3 spares = 27–30 turbines total.
Lead time problem: New GE Vernova units have a 3–5 year backlog. the applicant would likely need refurbished units or the LM6000VELOX fast-delivery variant.
Siemens & Mitsubishi
Less likely - why they're probably not the choice here
Siemens SGT-A35 (34–38 MW): Capable unit, acquired from Rolls-Royce. But Siemens reported zero aeroderivative orders in North America in 2023–2024. 3–5 year backlog same as GE. Packaged in Sweden - longer lead times for US projects.
Siemens SGT-800 (40–60 MW): Dominates the European market but has limited US data center footprint.
Mitsubishi FT8 (25 MW) / M501: FT8 saw a small resurgence in 2024, but Mitsubishi's aeroderivative market share is minimal in North America and declining. Their strength is in large-frame heavy-duty turbines (M501JAC, M701) at 300–500 MW - too large for modular behind-the-meter data center deployment.
Bottom line: GE Vernova holds ~95% of the North American aeroderivative market. The LM6000 or LM2500XPRESS is the near-certain choice.
Market data rules these out as primary candidates
Siemens SGT-800 (40–60 MW): Dominates the European market but has limited US data center footprint.
Mitsubishi FT8 (25 MW) / M501: FT8 saw a small resurgence in 2024, but Mitsubishi's aeroderivative market share is minimal in North America and declining. Their strength is in large-frame heavy-duty turbines (M501JAC, M701) at 300–500 MW - too large for modular behind-the-meter data center deployment.
Bottom line: GE Vernova holds ~95% of the North American aeroderivative market. The LM6000 or LM2500XPRESS is the near-certain choice.
Primary turbine comparison - what goes into Connaughtown's backyard
| GE LM6000 output per unit | 46–52 MW |
| Units needed for 850 MW (LM6000) | 18–21 turbines |
| GE LM2500XPRESS output per unit | 35 MW |
| Units needed for 850 MW (LM2500) | 27–30 turbines |
| Each unit footprint (LM6000) | ~4.9 m × 2.2 m engine |
| NOₓ emissions per unit | 15 ppm (DLE mode) |
| Combined NOₓ - 20 LM6000 units | ~2,000–4,000 tons/yr |
| CO₂ per unit (natural gas) | ~130,000 tons/yr each |
| Total CO₂ - 20 LM6000 units | ~2.6M tons/yr |
| New unit lead time (GE Vernova) | 3–5 years minimum |
| Refurbished unit lead time (ProEnergy) | Available 2027 |
Backup diesel generators - correcting the developer's claim
the applicant publicly cited 40 backup generators. This is not consistent with any known engineering standard for 850 MW.
The math: Standard data center backup gensets are 2–4 MW units. With N+1 redundancy for 850 MW:
• At 2 MW each: 425 units + 1 spare = 426 gensets minimum
• At 4 MW each: 213 units + 1 spare = 214 gensets minimum
• At 4 MW with 2N (full redundancy): 426 gensets
40 generators at 4 MW covers ~160 MW - less than 19% of the site load. A brownout under that configuration would be catastrophic for the data center and unacceptable for any hyperscale operator.
The only scenario where 40 is plausible: If "40" refers to turbine-class generators (e.g. 40 × 20 MW gas reciprocating engines = 800 MW) - but that is a very different technology class with far greater emissions, and still falls short of 850 MW.
Developer's 40-unit claim is off by a factor of 5–10×
The math: Standard data center backup gensets are 2–4 MW units. With N+1 redundancy for 850 MW:
• At 2 MW each: 425 units + 1 spare = 426 gensets minimum
• At 4 MW each: 213 units + 1 spare = 214 gensets minimum
• At 4 MW with 2N (full redundancy): 426 gensets
40 generators at 4 MW covers ~160 MW - less than 19% of the site load. A brownout under that configuration would be catastrophic for the data center and unacceptable for any hyperscale operator.
The only scenario where 40 is plausible: If "40" refers to turbine-class generators (e.g. 40 × 20 MW gas reciprocating engines = 800 MW) - but that is a very different technology class with far greater emissions, and still falls short of 850 MW.
What 18–21 LM6000 gas turbines running continuously means for Connaughtown
Sound
Each LM6000 enclosure emits 85–95 dB at 1 meter. With industrial silencing: ~65–75 dB at the enclosure boundary. 20 units running simultaneously produce a combined sound pressure exceeding any single unit - resonance and low-frequency harmonics from turbine exhaust stacks travel further than high-frequency noise and penetrate walls. Residents within 0.5 miles experience a constant industrial hum indistinguishable from day to night.
Exhaust stacks
Each LM6000 requires an exhaust stack typically 15-25 meters tall. 20 units = 20 visible industrial smoke stacks rising above the existing buildings. Neighboring property owners raised concerns about visible smokestacks at the October 2025 planning meeting. Hot exhaust plumes at 400-500°C rise and disperse NOx and CO2 over the surrounding community.
Air permitting
At 2,000–4,000 tons/yr of NOₓ, this facility would require a Title V Major Source Operating Permit and likely Prevention of Significant Deterioration (PSD) review - both multi-year processes under PA DEP. Given Montgomery County's near non-attainment status, emissions offset credits would be required. These are scarce and expensive in the PJM region. No permit has been applied for.
02
Power & Emissions
PECO / PJM Grid · PA rate 0.29 t CO₂/MWh
Power draw
850 MW
Always-on, every second
Annual energy
7.4 TWh
Per year
CO₂/year (PA grid)
2.1M tons
PA-specific emission factor
Avg PECO residential rate
16.6¢/kWh
25% above national avg
850 MW
Continuous power draw
= Philadelphia's entire peak residential load (~870 MW)
= 680,000 average Pennsylvania homes powered simultaneously
= 4× the peak load of Plymouth Township itself
Equivalent to a small city
= 680,000 average Pennsylvania homes powered simultaneously
= 4× the peak load of Plymouth Township itself
2.1M tons
CO₂ per year at PA grid rate
= 456,000 cars driven for one full year
= 10× all CO₂ from every Plymouth Township resident
= emissions of 290,000 US homes
PA grid - not renewables
= 10× all CO₂ from every Plymouth Township resident
= emissions of 290,000 US homes
Grid load comparison
03
Water & the Schuylkill River
0.4 miles from site · 2M downstream drinkers
Annual consumption
2.6B gal
Cooling tower evaporation
Daily draw
7.1M gal
Every single day
River distance
0.4 mi
Schuylkill at Plymouth Dam
Downstream population
2 million
Drink Schuylkill water in Philly
7.1M gal
Consumed daily - every day, forever
= daily water use of 56,000 people (Plymouth Twp. × 3)
= 10.8 Olympic swimming pools drained every single day
= 2× what Plymouth Township's entire water system delivers daily
Schuylkill watershed stress
= 10.8 Olympic swimming pools drained every single day
= 2× what Plymouth Township's entire water system delivers daily
+3–6°F
Thermal discharge into Schuylkill River
Kills cold-water fish species (American shad, river herring, smallmouth bass) within 0.5 mi of outfall. Dissolved oxygen drops 20–30% in affected stretch. Philadelphia Water Dept. intake is just 18 miles downstream. EPA 316(a) variance required - rarely denied.
Drinking water source threatened
Forever
PFAS contamination - does not break down
Many AI data centers use PFAS-based fluids for immersion cooling and AFFF fire suppression foam. PFAS contamination is linked to kidney and testicular cancer, thyroid disease, immune disruption, and reproductive harm. Once it reaches groundwater or the Schuylkill, remediation is essentially impossible.
Permanent groundwater risk
04
Air Quality & Emissions
Montgomery Co. already near EPA ozone non-attainment
~4,400 tons
NOₓ per year - smog & ozone precursor
= NOₓ from 220,000 cars driven one year
Montgomery County is already in partial EPA non-attainment for ozone. This facility pushes the region further into violation on hot summer days, triggering health alerts affecting all 864,000 county residents.
Montgomery Co. already impacted
Montgomery County is already in partial EPA non-attainment for ozone. This facility pushes the region further into violation on hot summer days, triggering health alerts affecting all 864,000 county residents.
~310 tons
PM2.5 fine particulates per year
Plymouth Meeting's current PM2.5 is already 6.9 µg/m³ - dangerously close to the EPA standard of 9 µg/m³. Grid-attributed particulates push this region toward non-attainment. Each 10 µg/m³ increase in PM2.5 is linked to a 0.9% rise in lung cancer mortality.
Near EPA limit already
215–430
Diesel backup generators required - developer claimed only 40
The proposal lists "ancillary standby generators" but the applicant publicly cited just 40 units. This is mathematically impossible for 850 MW. Industry-standard 2–4 MW diesel gensets with N+1 redundancy require 215–430 units. 40 generators at 4 MW each covers only ~160 MW - barely 19% of the load. The real number is 5–10× what was claimed. Diesel exhaust is an IARC Group 1 confirmed human carcinogen. Weekly load testing across hundreds of units = hours of simultaneous combustion equivalent to tens of thousands of diesel trucks idling - every week, forever.
IARC Group 1 carcinogen · Developer's count is wrong
05
Permanent Heat Dome
850M BTU/hr rejected to atmosphere - 24/7/365
850M BTU/hr
Continuous waste heat output
= heat output of 7,000 residential gas furnaces running simultaneously
= raises local air temperature 2–5°F within a 0.5-mile radius, permanently. This is not seasonal. It does not stop at night. It does not stop in winter.
Permanent - not seasonal
= raises local air temperature 2–5°F within a 0.5-mile radius, permanently. This is not seasonal. It does not stop at night. It does not stop in winter.
Summer temperature impact near site
+$180–$420
Added cooling costs per nearby home per year
Same heat island magnitude as downtown Philadelphia vs. its suburbs. Increases heat-related illness risk for outdoor workers, children, and elderly residents by ~35%. Those within 0.5 mi will run air conditioning longer and harder - at their own expense - to compensate for a facility they never consented to.
You pay for their heat
06
Noise - 24/7 Industrial Operation
PA DEP Chapter 147 - 50 dB nighttime residential limit
Noise levels by distance from 900 Conshohocken Rd
Fence line (50 ft)
75–85 dB
500 ft - nearest homes
55–65 dB
0.25 mile
47–53 dB
PA DEP night limit
50 dB limit
EPA school standard
45 dB standard
WHO sleep threshold
40 dB threshold
Documented health effects by noise level
| 40 dB - WHO threshold (night) | Sleep disruption begins |
| 45 dB - EPA school standard | Learning impairment risk |
| 50 dB - PA DEP residential limit | Stress hormones elevated |
| 55 dB chronic exposure | +14% hypertension risk |
| 60 dB chronic exposure | +17% heart disease risk |
| 65 dB chronic exposure | Cognitive decline, depression |
| 85 dB - generator tests | Hearing damage begins |
75–85 dB
At facility boundary - forever, no holidays
= standing next to a running lawnmower - permanently
= louder than highway traffic at 50 feet
= 3× the intensity of normal conversation
= nighttime level at 500 ft likely exceeds PA DEP limit
No seasonal break. No nights off.
= louder than highway traffic at 50 feet
= 3× the intensity of normal conversation
= nighttime level at 500 ft likely exceeds PA DEP limit
Colonial SD
Nearest school district to the site
Nearest elementary school: ~0.8 miles. WHO research links chronic noise above 50 dB to measurable reading delays and reduced working memory in children. Studies in Northern Virginia show data center hum penetrates classrooms 0.5 miles away. Acoustic retrofits cost ~$20,000 per classroom - paid by taxpayers.
Children's learning impacted
24/7 lighting
All-night exterior lighting - melatonin disruption
Blue-spectrum security lighting suppresses melatonin production in children and adults. Children are more sensitive - disruption documented at light levels as low as 10 lux (a streetlight). Sleep loss linked to attention deficits, lower academic performance, obesity, and increased illness in school-age children.
Melatonin disruption in children
07
Children's Health & Cancer Risk
Most vulnerable: children, elderly, pregnant women
3× higher
Children's PM2.5 respiratory sensitivity vs. adults (EPA)
Smaller airways, faster breathing rates, more hours outdoors. A 2025 modeling study (Tao & Gao, ScienceDirect) projects US data centers will cause ~600,000 asthma symptom cases and ~1,300 premature deaths per year by 2030. Plymouth Twp.'s PM2.5 is already near the EPA limit.
~600K asthma cases/yr nationally by 2030
OR 1.5–2.0×
Childhood leukemia odds ratio near high-voltage lines (9-study meta-analysis)
A 500kV substation required for 850 MW interconnection would be sited on or adjacent to this facility. IARC classifies ELF-EMFs as Group 2B (possibly carcinogenic). Fields exceeding 0.4 µT - the threshold associated with doubled leukemia risk - extend ~300 meters from transmission lines.
IARC Group 2B - possible carcinogen
Science contested. No confirmed causal mechanism. Risk is real but not proven.
Legionella
Cooling towers are primary Legionella amplification sites
850 MW requires massive wet evaporative cooling towers. Warm, aerosolized water droplets are ideal for Legionella pneumophila. If water treatment lapses, bacteria can travel up to 6 km downwind. Outbreaks have been traced to industrial cooling towers in densely populated areas. Plymouth Twp.: ~2,200 people/sq mi.
PA DEP registration required
PFAS
"Forever chemicals" - linked to cancer, immune damage, reproductive harm
PFAS-based cooling fluids and AFFF fire suppression foam are common in AI data centers. PFAS is linked to kidney cancer, testicular cancer, thyroid disease, and immune system damage. It does not break down in soil, water, or the human body. Contamination of the Schuylkill River would be effectively permanent - and Philadelphia's water intake is just 18 miles downstream.
Permanent contamination risk
Chronic disease pathways - noise alone
| Hypertension (chronic noise >50 dB) | +14% risk |
| Heart disease (night noise >40 dB) | +17% risk |
| Sleep disorders (within 0.25 mi) | Documented |
| Depression / anxiety | Documented |
| Cognitive decline (elderly) | Associated |
| Asthma exacerbation (PM2.5) | Strong evidence |
| Lung cancer (diesel PM) | IARC confirmed |
08
Wildlife & Ecosystem
Schuylkill Valley · Atlantic Flyway · Plymouth Creek watershed
500–800 acres
Permanent habitat destruction - site + substation + transmission
Habitat fragmentation is the #1 driver of species loss globally (IPBES). Construction permanently eliminates pollinator forage, nesting cover, and wildlife corridors connecting the Schuylkill River Greenway to inland habitats. The Schuylkill Greenway supports over 200 bird species, 40 fish species, and critical pollinator corridors for agricultural land to the north.
Permanent land conversion
"Sensory danger zone"
Designation for 24/7 lit, noisy industrial facilities (Univ. of Michigan research)
The Schuylkill River corridor is an active Atlantic Flyway migration route. All-night lighting attracts and disorients nocturnal migrants - causing window collisions and exhaustion mortality. Low-frequency noise disrupts bird communication and reproduction (barn swallow, wood thrush documented). Communication towers required kill ~7M birds/year nationally.
Atlantic Flyway impact
American shad
Keystone Schuylkill species threatened by thermal discharge
American shad, river herring, and smallmouth bass spawn near Plymouth Dam - all are critically temperature-sensitive. Each +1°C reduces dissolved oxygen ~2%. Cooling tower biocide discharge (chlorine dioxide, glutaraldehyde, copper-based algaecides) is lethal to macroinvertebrates and fish larvae. The Schuylkill has been recovering for 50 years - this reverses it.
50 years of recovery at risk
~40% decline
Insect biomass loss expected within habitat footprint
Light pollution kills and attracts moths, beetles, and other insects. Impervious surface eliminates 100% of ground-nesting bee habitat. Montgomery Co. pollinator corridors would be severed. Loss of pollinators cascades to reduced crop yields, reduced wild plant reproduction, and collapse of bird food chains.
Pollinator corridor severed
+40–60%
Stormwater runoff increase from impervious surface
Plymouth Township sits in a FEMA-mapped 100-year flood zone along the Schuylkill. This facility dramatically increases flash flooding risk to downstream neighborhoods. Hurricane Ida (2021), Irene (2011), and major floods in 2006 already devastated this corridor. Cooling chemical runoff contaminates stormwater entering the Schuylkill without full treatment.
FEMA flood zone - already at risk
Hazmat site
E-waste, batteries, refrigerants, solvents - ongoing hazardous waste stream
Server hardware contains lead, mercury, cadmium, beryllium, and brominated flame retardants. UPS batteries contain sulfuric acid and heavy metals. Hardware is replaced every 3–5 years - creating a permanent, large-scale hazardous waste operation next to a residential community and a drinking water river.
RCRA hazardous waste regulated
09
Your Utility Bills
Costs socialized across all PECO ratepayers
$9–$25
Added monthly cost per household
$108–$300
Added annual cost per household
$34M–$95M
Total annual burden on Montgomery Co. ratepayers
$1–3 billion
Transmission infrastructure upgrades required - paid by ratepayers
PJM requires a new 500kV substation for loads above 500 MW. These costs are filed as FERC-approved "necessary upgrades" and recovered from all PECO ratepayers - not the data center operator. Montgomery Co. already pays 25% above the national average ($176/month avg bill). This adds to a burden residents are already struggling with.
You pay. They profit.
What $1–3B in grid upgrades means
10
The $21 Million Tax Claim - Fact Check
Developer's own application said $5M · PA's 25-year tax exemption dramatically reduces the real number
Developer's public claim
$21M/yr
Disputed by board member
Developer's own filing said
~$5M/yr
Per Data Center Dynamics
PA sales tax exemption
25 years
All equipment purchases
PA statewide revenue loss
$2 billion
By 2031 (Spotlight PA, Feb 2026)
25-year exemption
Pennsylvania's Computer Data Center Equipment Tax Exemption Program (Act 84, 2016)
Under Pennsylvania law, certified data centers receive a 25-year sales and use tax exemption on ALL qualifying equipment - servers, power systems, cooling infrastructure, backup generators, and UPS systems. For an 850 MW facility, equipment costs run into the billions. The exemption on equipment purchases alone could shield hundreds of millions of dollars from state and local tax over the facility's life.
What this means for the $21M claim: Any tax revenue calculation that doesn't account for this exemption is inflated. The Shapiro administration estimates PA-wide data center tax exemption losses will reach $188 million in FY2026–27 and $517 million by FY2031 - more than the entire $21M claimed annually, statewide.
PA exemption cost: $2B by 2031 statewide (Spotlight PA)
What this means for the $21M claim: Any tax revenue calculation that doesn't account for this exemption is inflated. The Shapiro administration estimates PA-wide data center tax exemption losses will reach $188 million in FY2026–27 and $517 million by FY2031 - more than the entire $21M claimed annually, statewide.
LERTA + KOZ
Additional Pennsylvania tax reduction programs available to this developer
LERTA (Local Economic Revitalization Tax Assistance): Allows up to 10 years of property tax abatement on improvements to industrial properties, with abatement percentages declining by 10% per year. The former Cleveland-Cliffs mill site is exactly the type of property this program targets. If Plymouth Township grants LERTA, property tax revenue from the site could be near zero for a decade.
Keystone Opportunity Zone (KOZ): If the site were designated a KOZ, the effective tax burden could be reduced to zero across state and local taxes - property, income, and sales - for the zone's duration.
Federal bonus depreciation: Under the One Big Beautiful Bill Act (2025), 100% bonus depreciation is permanent for property placed in service after January 19, 2025, eliminating federal tax liability for years.
Combined incentives could reduce net tax to near zero for years
Keystone Opportunity Zone (KOZ): If the site were designated a KOZ, the effective tax burden could be reduced to zero across state and local taxes - property, income, and sales - for the zone's duration.
Federal bonus depreciation: Under the One Big Beautiful Bill Act (2025), 100% bonus depreciation is permanent for property placed in service after January 19, 2025, eliminating federal tax liability for years.
What other states show
Comparable hyperscale tax claims vs. reality - the pattern is consistent
Virginia: Home to the world's largest data center cluster. $732M in annual subsidies paid to the industry. Loudoun County hosts 25M+ sq ft of data centers - yet residents pay some of the highest electricity rates in the state and infrastructure costs are socialized broadly.
Georgia: Projected $327M in data center sales tax exemption losses for FY2025. Actual figure came in at $2.5 billion - nearly 8× the estimate (Atlanta Journal-Constitution). Underestimation is the industry norm, not the exception.
Good Jobs First (Princeton-affiliated research): Studies consistently show data center tax incentives "do not deliver broad long-term benefits" and provide poor return on investment for states and localities. Economic activity is capital-intensive, not labor-intensive - the $21M figure likely reflects gross assessed value, not net community benefit after incentives.
Georgia's estimate was off by 8× - PA's will be too
Georgia: Projected $327M in data center sales tax exemption losses for FY2025. Actual figure came in at $2.5 billion - nearly 8× the estimate (Atlanta Journal-Constitution). Underestimation is the industry norm, not the exception.
Good Jobs First (Princeton-affiliated research): Studies consistently show data center tax incentives "do not deliver broad long-term benefits" and provide poor return on investment for states and localities. Economic activity is capital-intensive, not labor-intensive - the $21M figure likely reflects gross assessed value, not net community benefit after incentives.
What $21M/yr actually buys Plymouth Township - if it materialized
| Developer's claimed annual tax revenue | $21,000,000 |
| Developer's own filing figure | ~$5,000,000 |
| Estimated net after PA 25-yr exemption | $3M–$8M/yr realistic |
| Annual ratepayer burden on Montgomery Co. | $34M–$95M/yr |
| Required grid upgrade cost (ratepayer-borne) | $1B–$3B total |
| PA sales tax exemption value over 25 years | $500M+ estimated |
| Net to community (tax minus grid costs) | Deeply negative |
Jobs: what hyperscale data centers actually create long-term
the applicant cited 12,000 union construction workers - a temporary figure during the build phase, spread over several years, likely across hundreds of contractors regionally.
Permanent operations jobs at hyperscale data centers: typically 50–200 people - security, maintenance, IT operations, and administration. Most are hired through local contractors, not the Big Tech tenant. Data centers are among the most capital-intensive, labor-minimal industrial facilities ever built.
Comparable hyperscale examples:
• Amazon's Bucks County PA facility (2M sq ft): ~200 permanent jobs projected
• Google's Virginia campuses (multi-GW): ~1,000 jobs across dozens of buildings
• Microsoft data center cluster, Quincy WA: 48 permanent employees for 500,000 sq ft
Good Jobs First research: "Data center jobs are often focused on security, maintenance, or administration - employed by local contractors, reducing job quality and reducing the direct benefit to the community."
12,000 workers = temporary construction, not permanent jobs
Permanent operations jobs at hyperscale data centers: typically 50–200 people - security, maintenance, IT operations, and administration. Most are hired through local contractors, not the Big Tech tenant. Data centers are among the most capital-intensive, labor-minimal industrial facilities ever built.
Comparable hyperscale examples:
• Amazon's Bucks County PA facility (2M sq ft): ~200 permanent jobs projected
• Google's Virginia campuses (multi-GW): ~1,000 jobs across dozens of buildings
• Microsoft data center cluster, Quincy WA: 48 permanent employees for 500,000 sq ft
Good Jobs First research: "Data center jobs are often focused on security, maintenance, or administration - employed by local contractors, reducing job quality and reducing the direct benefit to the community."
11
The math doesn't add up for Plymouth Township
Community costs
| Residents bearing noise/heat/air impacts | ~18,500 people |
| Households facing utility bill increase | ~7,200 homes |
| Children in affected school district | ~4,000 children |
| Downstream drinking water risk | 2 million people |
| Homes losing $27K–$81K in equity | Entire 0.5 mi radius |
| Annual ratepayer burden, Montgomery Co. | $34M–$95M/yr |
| Required grid infrastructure cost | $1–3 billion |
What the community gets
| Permanent local jobs created | ~50–200 |
| Construction jobs (temporary, years) | ~1,000–2,000 |
| Annual tax revenue to township | $5M–$15M est. |