A 250 MW data-center campus sited in Lycoming County under this ordinance writes the host township and county a check for roughly $8–12 million every year — on top of a $20M up-front community trust, a fully-funded fire engine, dark fiber to every school, and a Penn College pipeline that keeps our kids home. The fuel is ours. The water is ours. The terms are ours. This is what a serious host community asks for.
Every figure below is a contractually enforceable obligation under this ordinance — not a pitch deck, not a press release, not “voluntary community engagement.” Most are written directly into zoning. The rest are conditions of conditional-use approval, executed before a shovel hits ground. Walk away from any one of them and the project doesn’t get built.
Plus: a new Class A fire engine and lithium-fire response trailer for the host VFC; annual specialized training (Class C, lithium thermal runaway, gas-turbine response, HV substation rescue) at the developer’s expense; a public dashboard for noise, water, and energy use; and a permitted-reuse first-option clause that gives the buildings a second life. Every commitment is written into the ordinance — section by section →
The 2026 PA DEP Citizens Advisory Council guide and most county-level templates were drafted to be politically palatable to developers. Useful as a baseline. Insufficient as a deal. Lycoming’s ordinance pulls the same structural bones and then sharpens every clause where the host community has actual leverage.
None of these provisions kill the project economics. Hyperscalers routinely build under stricter regimes in northern Virginia, Ireland, and Singapore. What this ordinance does is make sure the value created in the valley stays in the valley.
Lycoming County is not in crisis. It is in slow decline — the kind that doesn’t make headlines, but quietly hollows out main streets, school districts, and tax rolls. Median household income trails the state by roughly $14,000. Population has fallen every census since 1980. Williamsport raised property taxes again in 2025 to plug a budget gap, despite a 92% collection rate. The trajectory is fixable. The window is shorter than people think.
Hyperscalers are not looking for the cheapest land. They’re looking for the rare combination of cheap, reliable power within transmission distance, sufficient water, a workforce that won’t flee, and a county government that can issue a permit before the project economics rot. Lycoming has all five.
The 870 MW Hamilton Patriot combined-cycle plant in Clinton Township is already connected to PJM. Pennsylvania sits atop the Marcellus, producing enough gas to power the entire eastern interconnect for decades. New PJM capacity auctions are clearing at record prices — generation-adjacent siting is now worth real money to a hyperscaler. The same Marcellus pipeline grid that feeds the plant feeds the backup generators too: this ordinance requires natural-gas standby units in place of diesel, so even emergency power comes off Pennsylvania’s own gas.
The West Branch of the Susquehanna passes through the county at sustained, predictable flows — none of the aquifer stress that has stalled projects in Loudoun County or Phoenix. Closed-loop cooling pulls a fraction of what data-center critics fear, and a strong ordinance can require it.
Pennsylvania College of Technology graduates electricians, HVAC technicians, mechatronics engineers, and welders by the thousand — most of them currently leaving for jobs out of state. A 100 MW data center campus needs 200–300 of those people permanently, at wages that compete with Marcellus drilling.
The ridge-and-valley topography of north central PA gives developers what coastal sites can’t: rock at shallow depth, low seismic risk, no flood-plain federalism, no hurricane insurance premium. Construction costs run 20–30% below the Virginia data-center alley average.
The Pennsylvania Computer Data Center Equipment Exemption Program waives state sales tax on servers, cooling, and software for projects investing ≥$75 million and creating ≥25 jobs. House Bill 2151 (April 2026) directs DCED to publish a statewide model ordinance — Lycoming can be the first county in line, not the last.
A typical 300-home development in this part of Pennsylvania generates about $17 million in tax revenue over seven years. A single mid-sized data center generates roughly $93 million over the same period — without sending one new student to the school district.From a Pennsylvania county fiscal analysis cited in industry briefings on PA data-center buildout
The Pennsylvania DEP Citizens Advisory Council published a Data Center Ordinance Guide in 2026 — a 30-page toolkit municipalities can copy almost verbatim. Below is a synthesis of its core provisions, tuned for Lycoming County conditions. Every number is defensible. Every standard is enforceable. Read the full ordinance text →
Permit data centers only in I‑1 / I‑2 industrial districts with a minimum 5‑acre lot for any facility exceeding 100,000 sq ft. Require 1,000 feet of separation from any residential use or sensitive receptor, reducible to 500–999 feet only if the applicant exceeds ambient‑noise and visual‑screening standards.
Cap sound at the property line at 40 dB(A) / 50 dB(C) overnight (7 pm–7 am) and 45 dB(A) / 60 dB(C) during the day. Require pre-construction and post-occupancy noise studies, conducted between one and twelve months after the first Certificate of Occupancy.
Require closed-loop cooling unless an alternative demonstrably uses less water and less energy. Prohibit private wells or stream withdrawals as a primary cooling source when public water is available. Require a 1-mile-radius hydrogeologic study, drought response plan, and PADEP / SRBC permits before approval.
Primary backup generators shall be lean-burn natural gas, fueled via firm-contract Marcellus pipeline supply — not diesel. Gas engines meet EPA Tier 4 without aftertreatment, cut NOx and particulates dramatically, and run on the fuel we already produce. Diesel is permitted only as a fuel-of-last-resort for pipeline outages, capped at 20% of backup capacity and 72 hours of on-site fuel. No peak shaving. No grid arbitrage. Annual emissions reports to the municipality.
Pair the state sales-tax exemption with an aggressive LERTA‑style local abatement schedule that starts at full-tax assessment of real estate from year one. A 100 MW facility paying Lycoming property tax at current millage rates yields roughly $4–7M annually to the county, school district, and city — before any community benefit agreement.
Require a financial surety equal to 125% of demolition and site-restoration cost, CPI-escalated. If data-center use ends, the owner has 12 months to either pursue a permitted reuse, market the property as industrial reuse, or begin decommissioning. Regulated materials (refrigerants, batteries, fuels) come out within 180 days regardless. Buildings worth keeping can stay. No bond, no permit.
Per PA Act 247, Community Benefit Agreements can’t be codified in zoning — but they can be negotiated as a separate parallel contract before land development submission. Target: emergency-response training and reimbursement, workforce-training trust for Penn College, dual-enrollment cloud-computing curriculum, and a public dashboard for noise, water, and energy use.
Compare the fiscal footprint of one mid-sized hyperscale campus against the most-typical alternative development scenarios. The data center costs the county effectively zero in new services — no school enrollment, marginal road wear, no expansion of social services — while paying property tax on a $1.5–2 billion improvement basis.
Opposition to data centers is loud right now — in Bucks County, in northern Virginia, in Memphis — and the reason is almost always the same: residents see the wires going in and don’t see anything coming back. Lycoming’s ordinance answers that objection on page one. Every megawatt of approved load triggers a stack of legally enforceable obligations to the host township, the county, the school district, and the airport. This is not a wish list. Every mechanism below is permitted under existing Pennsylvania statutes and has been used somewhere in the Commonwealth or in a peer state.
Authority: Act 511 (Local Tax Enabling Act) + negotiated host agreement, mirroring the PA host municipality benefit fee model used for waste facilities since 1988.
A 250 MW campus pays the host township $875,000/year in unrestricted general-fund revenue — escalated 2.5% annually, paid quarterly, not abated. For Old Lycoming or Armstrong Township, that single line item is larger than the entire current police budget.
Authority: Pennsylvania LERTA (Act 76 of 1977) restructured as a Payment In Lieu Of Taxes agreement.
Developers will ask for a LERTA tax abatement — NorthPoint already did exactly that in Hazle Township, Luzerne County. Lycoming’s ordinance says no to abatement and yes to a PILOT: the developer pays a contractually fixed annual amount equal to at least 80% of the full-build property tax, escalated, with revenue split 60% school district / 30% township / 10% county. The school district gets a predictable, dedicated revenue stream the first day construction starts — not in year eleven.
Authority: Act 511 BPT, codified by Act 42 of 2014’s 15-day nexus rule.
Data centers are a service business under Pennsylvania law — they sell compute, storage, and bandwidth — which means they are not shielded by the manufacturing exemption that protects Marcellus producers. The host township levies BPT at 1.5 mills on in-jurisdiction gross receipts. On a $400M-revenue tenant, that is $600,000/year the township keeps.
Dedicated, voluntary contribution in the host agreement; modeled on the Surf Air – Williamsport community sponsorship program already in use at IPT.
Williamsport Regional Airport lost American Airlines and is now served by a single 9-seat Cessna Caravan operated by Southern Airways Express to Dulles — propped up by a federal EAS subsidy that has climbed past $4 million per year. A real data-center campus brings exactly the corporate travel demand needed to support a CRJ-200 or ERJ-145. Three facilities at $250k each fully fund the local cash match the 2019 EAS proposal required and add a marketing reserve. Data centers literally pay to bring jet service back to IPT.
Negotiated CBA contribution; Penn College Workforce Development already runs custom industry training contracts at scale.
Each developer endows a Data Center Operations & Mechatronics track at the Pennsylvania College of Technology: full tuition for 20 Lycoming and Clinton County residents per cohort, paid apprenticeships during build-out, and a guaranteed interview at the host facility. This is the single most powerful tool to keep young people from leaving the valley — and it turns Penn College from a regional asset into a national pipeline.
In-kind contribution requirement, enforceable as a condition of conditional-use approval under the PA Municipalities Planning Code.
Data centers run massive dark-fiber rings. The ordinance requires every developer to extend 4 strands of dark fiber, at cost, to: the host township building, the nearest fire/EMS station, the nearest public school, and one designated rural anchor (library, senior center, or municipal park). Recurring transport is billed at the regulated wholesale rate. Rural Pennsylvania gets symmetric gigabit backhaul as a byproduct.
Conditional-use exaction tied to documented service-call impact; permitted under PA MPC § 503-A.
A 250 MW campus generates one $125,000 capital contribution to the host volunteer fire company — enough for a new Class A engine or a fully-equipped lithium-fire response trailer — plus $37,500/year recurring for training and SCBA replacement. The host VFC stops bingo-night fundraising for trucks.
Modeled on the NAACP Community Benefit Agreement template (2026), with a citizen-led Community Advisory Board governing disbursement.
A $2B facility seeds a $20M trust at construction commencement, with the principal held by the Lycoming County Community Foundation and disbursed against a published annual plan. Eligible uses: housing rehab, recreation, riverfront access, mental-health services, public-school capital. The board has resident, township, and developer seats — with the developer in the minority.
That is the difference between a township that tolerates a data center and a township that fights to host one.
None of the numbers below are projections. Each is the documented, current fiscal reality of a county or town that wrote a real ordinance, signed a real agreement, and is now living on the proceeds. Lycoming is not late to this — it is just early enough to learn what to insist on.
The benchmark. 4% of land. 45% of revenue.
A 7,500-person farm town. A new high school.
Post-textile collapse. Google as anchor tenant.
12,000 residents. 560 permanent technical jobs. $490M/yr.
The lesson from all four is the same: the communities that insisted on real terms — PILOTs over abatements, host fees, dedicated school revenue, workforce pipelines, in-kind infrastructure — came out ahead. The communities that took whatever the developer offered did not. Lycoming’s ordinance is written from the first group’s playbook.
The opposition rarely asks what the compute is for. It is worth answering directly. Below are eight production deployments — not pilots, not promises — where the AI workloads running inside data centers are delivering measurable, peer-reviewed wins for human and planetary health. Every one of them runs on the kind of facility this ordinance regulates.
50% water savings · 20–30% yield gain
AI-driven irrigation systems combine soil-moisture sensors, hyperspectral drone imagery, and weather models to deliver water only where and when crops need it. A 2025 systematic review and meta-analysis across dozens of deployments documents 30–50% water savings and 20–30% productivity increases over traditional irrigation — the largest near-term agricultural water-conservation lever currently available. Agriculture uses roughly 70% of all freshwater withdrawals worldwide.
Source: AI-driven Irrigation Systems: Systematic Review & Meta-Analysis (2025); UCL deep-learning water-demand model, Scientific Reports (Oct 2024)
350,000 gallons/day recovered · $213K/yr saved
A Midwest water utility had spent two years trying to find a single phantom leak with three different traditional techniques. CivilSense AI, trained on 2.3 million acoustic signatures, pinpointed it: a 1/16-inch circumferential break on a 6-inch main, leaking nearly a quarter-million gallons per day into a storm drain 334 feet away. South Carolina’s Greenville Water saved an estimated 71 million gallons annually using the same class of system. Sweden’s VA SYD utility used Siemens SIWA Leak Finder to detect leaks as small as 0.5 L/sec and cut non-revenue water from 10% to under 8%.
Source: Oldcastle Infrastructure case study (Jun 2025); Greenville Water + VODA.ai; Siemens VA SYD case study
7.5M tons/yr wasted methane found
EDF’s MethaneAIR and MethaneSAT instruments, processed through a custom AI pipeline developed with Harvard and the Smithsonian Astrophysical Observatory, measured methane plumes across more than 70% of U.S. onshore oil & gas production. The verdict: actual emissions ran over four times higher than EPA estimates — enough wasted gas to heat half of all U.S. homes. The data, freely released, is now driving binding methane reduction commitments globally toward the 75%-by-2030 oil-and-gas target.
Source: EDF MethaneAIR program; World Economic Forum (Jul 2025)
10-day global forecast · minutes, not hours
DeepMind’s GraphCast runs a 10-day global weather forecast on a single TPU in under a minute — replacing supercomputer ensembles that take hours. It predicted Hurricane Lee’s Nova Scotia landfall nine days in advance, far ahead of conventional ensembles. ECMWF, the European weather agency, now runs GraphCast operationally. Earlier and more accurate forecasts translate directly to evacuation lead time, fewer storm deaths, and billions in avoided property loss.
Source: Google DeepMind GraphCast; AI Weather Forecasting 2026 review
+98.4 mL FVC vs. −20.3 mL placebo
Insilico Medicine’s rentosertib, a TNIK inhibitor for idiopathic pulmonary fibrosis, is the first drug for which both the target and the molecule were discovered entirely with generative AI — and the first to clear a Phase IIa trial. Published in Nature Medicine (Jun 2025): patients on 60 mg daily gained 98.4 mL of forced vital capacity; placebo patients lost 20.3 mL. Designed in 18 months for roughly $150,000 versus the hundreds of millions and decade-plus of traditional pharma. IPF was previously a slow death sentence.
Source: Insilico Medicine 2025 Annual Results; AI Agents in Pharma (Apr 2026)
2024 Nobel Prize in Chemistry
For 50 years, the protein-folding problem was biology’s grand challenge: knowing a protein’s shape is the prerequisite to designing a drug against it. Before AlphaFold, only ~48% of the human proteome had structural coverage. After: 76%. Human proteins with no structural information at all dropped from 5,027 to 29. The AlphaFold Protein Structure Database now holds over 200 million predicted structures, used in active drug-design work against malaria, tuberculosis, Chagas, cancer, prion diseases, and Alzheimer’s. AlphaFold3 reached 76.4% accuracy in protein-ligand docking — a 1.8× jump over prior methods.
Source: AlphaFold 3 transformative impact, Frontiers in AI (Apr 2026); Biolife Health AlphaFold review (Jun 2025)
~20% more cancers detected in real-world trials
A nationwide real-world study published in Nature Medicine (Jan 2025) and a 2026 randomized controlled trial both found that AI-assisted mammography reads detect more clinically relevant breast cancers than radiologists alone, while keeping the false-positive rate flat. This is detection at the population scale — the screening protocol seen by tens of millions of women per year — not a research curiosity. Cancers caught earlier mean lumpectomies instead of mastectomies, and survival instead of palliation.
Source: Nationwide real-world AI cancer detection, Nature Medicine (Jan 2025); The Guardian (Jan 2025)
10× bacterial reduction in mouse models
MIT’s Collins lab used deep-learning models to screen 39,000 compounds for activity against Acinetobacter baumannii, one of the WHO’s highest-priority drug-resistant pathogens, and identified abaucin — an entirely new structural class. A follow-up 2023 effort screened millions of compounds for activity against MRSA, surfacing two new structural classes that reduced bacterial populations 10-fold in mouse models without toxicity to human cells. Antibiotic resistance is projected to cause 10 million deaths per year by 2050. Traditional discovery has been nearly dead for two decades; AI just restarted the pipeline.
Source: MIT News — abaucin discovery (May 2023); MIT News — new MRSA antibiotic class (Dec 2023)
None of this happens without the compute. AlphaFold’s training run consumed hundreds of TPU-years. GraphCast trains on decades of reanalysis data. Insilico’s generative chemistry pipeline runs continuously across thousands of GPUs. Every breast-cancer screening AI deployed in a hospital was trained on a data-center cluster somewhere, and inferences against it run in a data center every time a radiologist clicks “analyze.” The buildings this ordinance regulates are the physical infrastructure of the most consequential scientific work of the decade. Lycoming County has the chance to host some of it, on terms it sets, with the dividends going to its own schools, fire halls, and main streets.
Forget the buzzwords. Here is what the work in those four healthcare cards above means for real people — a neighbor, a parent, a kid — explained the way you’d explain it across a kitchen table.
It’s called idiopathic pulmonary fibrosis — IPF for short. Doctors don’t know what causes it. The lungs slowly turn into scar tissue. The person literally can’t breathe.
A team at a company called Insilico used AI to do two things no human team had pulled off together. First, the AI read through millions of pages of biology research and figured out which switch inside the lung was probably broken. Then a second AI designed a brand-new molecule shaped exactly right to flip that switch back.
They ran the drug through a real human trial. Published the results in Nature Medicine in June 2025. The patients on the drug actually gained lung function. The placebo group kept losing it. That is the first time anyone has reversed a measurement of this disease with a pill.
Total cost to design the molecule: roughly $150,000. Total time: 18 months. A traditional pharma company would have spent ten years and hundreds of millions of dollars — and most of the time, ended up with nothing. The compute that made it possible runs in buildings exactly like the ones this ordinance regulates.
Everything alive is built out of tiny machines called proteins. Your muscles, your immune system, the stuff that reads your DNA — all proteins. They are folded into very specific shapes, and the shape is what makes them work.
An AI called AlphaFold, built by a Google team in London, taught itself the rules for how proteins fold. Then it predicted the shapes of every protein we know about — all 200 million of them — in about a year. It put the whole library online. For free. The scientists who built it won the Nobel Prize in Chemistry in 2024.
What does that get us? Drug research that used to take a decade now takes months. There are active programs right now using AlphaFold to design treatments for malaria, tuberculosis, Alzheimer’s, prion diseases, and pancreatic cancer. Real labs. Real patients. Real shots on goal that we did not have five years ago.
The training run for AlphaFold consumed the equivalent of hundreds of years of computer time on the most expensive chips made. It happened in a data center.
A mammogram is the X-ray a woman gets every year or two to check for breast cancer. A radiologist looks at hundreds of them a day. They’re good. They’re also human. They get tired.
A real-world study published in Nature Medicine in January 2025 ran the comparison across hundreds of thousands of women. A randomized clinical trial in 2026 confirmed it. The AI caught roughly 20% more real cancers than the radiologist working alone. And the false-alarm rate did not go up. Same number of unnecessary biopsies. More actual tumors found.
Why that matters in dollars and lives: cancer caught at stage 1 means a lumpectomy and the woman keeps her breast and goes back to work. Cancer caught at stage 3 means chemotherapy, a mastectomy, and a coin-flip on survival. The math turns into mothers who watch their kids graduate high school. Every single screening read by these AIs is processed in a data center the second the radiologist hits a button.
Some bacteria have figured out how to shrug off our antibiotics. The famous one is MRSA — a staph infection that doesn’t respond to penicillin or anything in its family. You can pick it up in a hospital from a knee surgery or in the locker room from a cut.
A lab at MIT pointed an AI at a chemical library of millions of possible molecules and asked it: which of these could kill the bacteria that humans have run out of ways to fight? The AI flagged a handful. The scientists made them in the lab. Tested them.
Two entirely new classes of antibiotic that work against MRSA. One that works against another nightmare bug called Acinetobacter baumannii — the World Health Organization’s number-one priority pathogen. All three killed the bacteria in mice without poisoning the mice. None of these existed before AI was pointed at the problem. The screening run that found them used the kind of GPU clusters that live in — you guessed it — data centers.
None of these are research curiosities. The IPF drug is in late-stage human trials. AlphaFold is used by every major pharma company on Earth. AI mammogram reads are now standard in dozens of European countries and U.S. hospital systems. The new antibiotics are headed to clinical trials. Every one of them was made possible by compute infrastructure that has to be built somewhere. Lycoming County can be one of those somewheres — and write the rules so the benefits stay home.
A data center is a big building full of computers. So how does it help the guy who runs the diner in Montoursville, or the lady who teaches third grade in Hughesville, or the kid who wants to apprentice at the electrical union hall? Here’s the path the money takes — step by step — using the numbers the U.S. Chamber of Commerce and the Commonwealth of Virginia have already published.
Above: typical impact of one mid-sized (165,000 sq ft) data center build, per the U.S. Chamber of Commerce data center economic impact study. After construction, the same facility supports 157 ongoing jobs, $7.8 million in annual wages, $32.5 million in annual local economic output, and $1.1 million per year in continuing state and local tax revenue.
He’s an IBEW Local 812 member from Williamsport. The contract pays union scale for 18 months. He pays off his truck. He keeps his health insurance. His kid stays on the family plan.
Groceries at Weis. Breakfast at the Bullfrog Brewery. New tires at the shop on Route 15. The lumber yard sells him deck boards. The dentist on Memorial Avenue finally gets her crown order paid.
The diner adds a second cook. The tire shop puts on another mechanic. The lumber yard keeps the third truck on the road. Each of those people now has a paycheck. They spend it too.
The facility itself pays property tax and a negotiated Host Community Agreement (see Pillar 02). That money goes to the school district, the volunteer fire halls in Muncy and Hughesville, the bridge fund, and the EMS that shows up when your father has a heart attack.
The Commonwealth of Virginia — which now hosts more data centers than any other state — commissioned Mangum Economics to count the actual jobs in 2025. The verdict: for every one job inside a data center, 2.1 more jobs exist outside it across the rest of the state economy. Construction, equipment supply, restaurants, truck repair, daycare, accounting, dentists. The whole web.
Statewide impact: $40 billion in economic output, 112,000 jobs, $9 billion in worker pay, and $1.5 billion a year in state tax revenue — from data centers alone. Source: 2026 NVTC Data Center Report, announcement here.
Loudoun County, Virginia—the largest data center cluster on Earth—ran the math. For every dollar the county spends serving a data center (roads, fire response, permits), the county receives $26 back in property and equipment taxes. A manufacturing plant returns about $4. A residential subdivision costs the county money — the kids cost more to educate than the property tax brings in.
In Lycoming County, that ratio funds the school district that lost enrollment to the gas patch boom-and-bust. It funds the volunteer fire halls that keep losing volunteers. It builds the sidewalks Williamsport hasn’t been able to afford since the airline left.
Penn College of Technology in Williamsport already trains electricians, HVAC techs, and welders. A signed data-center pipeline means three-year apprenticeships with guaranteed placement — the kind of work a high-school graduate can start at 19 and own a house by 25.
Williamsport Area, East Lycoming, Loyalsock, Montoursville Area, and Muncy school districts all share the county tax base. A 200-MW campus generates more tax revenue than 1,000 new homes — without adding 1,000 new students to feed and educate.
The Host Community Agreement (Pillar 02) lets the township negotiate dedicated payments to Muncy, Hughesville, Montgomery, and Picture Rocks fire companies — the same volunteers who would respond to a data-center incident anyway, now properly equipped to do it.
Eighteen months of 1,500 construction workers eating lunch in Williamsport, Muncy, and Montoursville. Then 150 well-paid permanent staff doing the same for 20 years. That is the math that keeps the Bullfrog, the Old Corner, the Beiter’s, and a dozen others off the closed-for-good list.
Permanent operations staff are W-2 employees with employer health insurance. That’s billable visits at UPMC Williamsport and Geisinger that replace the unreimbursed-care load the county has carried since the gas boom faded.
Per Pillar 04, primary power runs on natural gas turbines burning Marcellus methane from wells in Lycoming County itself. That’s royalty payments to local landowners, lease payments to the State Forest, and severance taxes that stay in Pennsylvania instead of flowing to whoever buys the LNG export shipment.
Industry brochures sometimes claim “multipliers” as high as 7 or 8 jobs per direct job. That number comes from one PwC study and economists like the University of Texas’s Nathan Jensen consider it inflated. The 2.1× figure above is the Commonwealth of Virginia’s own number from its 2026 report — produced by an independent economic firm, audited by a state body, and reflecting actual jobs counted, not modeled. We’d rather start with a number you can trust than one a developer would prefer.
A good ordinance is not built by ignoring opposition. It’s built by writing the opposition’s legitimate concerns into the regulatory text — so the bad actors get filtered out before they ever apply.
Not under this ordinance. The model language requires closed-loop cooling by default, prohibits private wells as a primary cooling source when public water is available, and forces a 1-mile-radius hydrogeologic study before any approval. A 100 MW closed-loop facility uses roughly the water of a mid-sized hotel — a fraction of the Loudoun County, Virginia evaporative-cooling designs that have produced the worst headlines.
The Susquehanna River Basin Commission has explicit, enforceable rules on consumptive use. Any applicant who can’t pass an SRBC water-feasibility study can’t pass this ordinance either.
PJM capacity prices are rising — but the cause is grid-wide demand growth, not whether any specific data center lands in Pennsylvania or somewhere else. Lycoming is already inside PJM. If the load lands here, the local tax base captures the upside. If it lands in Ohio, Pennsylvanians pay the higher capacity price and get none of the offsetting tax revenue.
The ordinance also prohibits generators from being used for peak shaving or grid arbitrage — closing the most-criticized loophole in other states’ rules.
Per dollar of capital invested, data centers create fewer jobs than light manufacturing. Per dollar of local tax revenue generated, they create dramatically more. The honest framing: a 100 MW campus brings ~200–300 permanent jobs at $75k–$120k median wages, plus 1,500–2,500 construction-phase jobs, plus a tax base that funds schools and roads without enrolling a single new student.
For a county that has been losing 100–200 working-age adults per year for two decades, that is not a small thing.
The 40 dB(A) overnight property-line cap in this ordinance is stricter than EPA’s recommended outdoor residential limit of 55 dB(A) day-night average. It is met routinely at modern facilities — through sound walls, equipment enclosures, and 1,000-foot setbacks. The 100-hour annual cap on generator testing, restricted to 9 am–3 pm weekdays, addresses the second most-common complaint.
The gas-primary requirement in § 5.1 helps further: lean-burn natural gas gen sets typically run 3–5 dB(A) quieter than equivalent-kW Tier 4 diesels at the property line, and they idle without the characteristic low-frequency diesel knock that carries furthest into residential areas.
Not in Lycoming. The ordinance flips the industry default: standby generators are natural gas, not diesel. Lean-burn gas engines meet EPA Tier 4 emission limits without the SCR and particulate filters diesels need — cutting NOx 50–70%, eliminating visible particulate plume, and lowering CO₂ per kWh by 20–40% (Pacifico Energy).
Diesel is permitted only as a fuel-of-last-resort for the rare case where the gas pipeline itself is down — capped at 20% of backup capacity, 72 hours of on-site fuel, 50 test hours per year, and Tier 4 final with aftertreatment. Caterpillar’s Fast Response gas gen sets (Oct 2025) hit cold-start in 6.5–7.5 seconds with full block-load acceptance — competitive with diesel and NFPA 110 Level 1 Type 10 certified.
For a county whose economy already runs on Marcellus gas, this is the obvious move: keep the fuel local, keep the air clean, and skip the diesel tank farm.
The ordinance plans for that in two layers. First, the buildings themselves are typically valuable to a successor industrial, logistics, light-manufacturing, R&D, or institutional user — they come with abundant power service, fiber, water, and stormwater infrastructure already built. Pennsylvania industrial property routinely transitions between uses, and § 7 explicitly favors adaptive reuse over knock-down. The owner has 12 months to pursue a permitted reuse or actively market the property for one, with up to three years for the marketing pathway.
Second, in case reuse never materializes, the ordinance requires a financial surety bond at 125% of demolition cost, CPI-escalated annually — no bond, no permit. Regulated materials (refrigerants, lithium-ion batteries, fuels, dielectric fluids) come out within 180 days regardless of reuse plans. The county will not be left holding a derelict slab full of hazardous fluids — that risk lives with the operator.
By contrast, no such bond protects the county from an abandoned warehouse, a defunct mall, or the bankrupt natural gas operators of the 2010s.
The state sales-tax exemption is set by Harrisburg, not the county — it applies whether the facility lands in Lycoming or somewhere else. The local question is narrower and starker: do we collect property tax on a $1.5B improvement in our county, or in someone else’s?
The model ordinance does not recommend a local property-tax abatement. It assesses real estate at full value from year one. The state subsidy is on the equipment inside the building. The local government keeps the building itself on the rolls.
The next 12 months will set Pennsylvania’s data-center map for a decade. Bucks, Berks, and Beaver counties are already negotiating. Lycoming should not be the county that finds out too late.
Every claim on this site links to a primary source. Civic policy needs to survive citation.