Ohio EPA at epa.ohio.gov. Division of Air Pollution Control permits are searchable. Permit-to-Install and Permit-to-Operate are the two relevant categories.
Other states. Search “[state] air permit search” or “[state] DEQ public notice” or “[state] DEP air permit.” Every state has it. The interface varies. The data does not.
What to look for. Two things. First, whether a permit application exists for the project at all. Second, whose name is on it. The applicant identity is one of the most informative pieces of public data for understanding how the deal is actually structured.
A permit filed under the data center operator’s legal entity name signals a direct-build, where the developer or operator owns the on-site generation and the math runs through the developer’s own books.
A permit filed under a utility name signals a regulated-utility build, where the utility owns the plant and recovers the cost through its rate base. A permit filed under an independent power producer or a separate generation LLC signals a merchant or contracted-power structure, with a separate operator selling the output back to the data center under a power purchase agreement. Each structure produces a different answer to who pays what when the contract expires.
The second thing to look for is the nonattainment overlay. In counties designated as nonattainment for any criteria pollutant, the application triggers Nonattainment New Source Review (NNSR). NNSR requires emission offsets, a more demanding Lowest Achievable Emission Rate analysis instead of the looser Best Available Control Technology standard used in attainment areas, and longer agency review. Developers rarely disclose NNSR exposure in their hearing presentations because the offsets are expensive and sometimes unavailable in thin local markets, and the longer review changes the project’s economics and timeline.
The third use of the air permit is as a description of the equipment. The application narrates the generation technology in plain language: gas turbine make and model, generator class, peak power output, chiller plant capacity, cooling architecture. The community can read the cooling design off the air permit even when the developer’s hearing slide hedges on which architecture is “currently planned.” If the slide says “evaluating closed-loop” and the permit application describes evaporative cooling, the permit application is the binding statement of what is actually being built.
MATH
Georgia Board: Bottom Line
The Georgia Department of Audits stated explicitly in the 2023 review that the exemption’s cost-effectiveness was difficult to assess and that the foregone revenue substantially exceeded the documented economic benefit.
Virginia: Bottom Line
Virginia JLARC reached similar conclusions in 2024 on Virginia’s incentive program. Both auditors are nonpartisan state offices reporting to legislatures of mixed control. Both reports are public. Most county commissioners voting on data center abatements have not read either.
10. “We will be operational in 18 months.” Sometimes 24.
The truth. From conditional use approval to power-on, even on the most aggressive realistic timelines, hyperscale takes 36 to 60 months. The constraints are physical. Large power transformers ran 128 to 160 weeks of lead time through 2025 and have not normalized. Medium-voltage switchgear runs 40 to 60 weeks. Caterpillar G3520K reciprocating generators were booking into 2029 in early 2026. PJM cluster studies run 18 to 36 months and the deliverability assignment is the gate to construction. Interconnection upgrades on the utility side are not on the developer’s schedule. Environmental review can add 6 to 12 months.
Source. PJM interconnection queue documentation. ERCOT generator interconnection queue. EPRI and Wood Mackenzie reports on grid equipment lead times.
Counter-question. “Provide the developer’s critical path schedule, by phase, with the long-lead equipment delivery dates, the interconnection study completion date, the utility-side upgrade construction schedule, and the substation energization date. Does the developer hold signed purchase orders (not quotes) on the long-lead equipment?”
11. “The facility’s air quality impact is negligible. Backup generators are tested infrequently.”
The truth. In ozone nonattainment areas, backup diesel generators alone can push a facility over New Source Review thresholds. The 25 tons-per-year nitrogen oxides threshold applies in Severe ozone nonattainment areas: the Philadelphia 5-county area, the Houston-Galveston-Brazoria area, the Dallas-Fort Worth 10-county area, parts of the Northeast Corridor, the LA Basin. NSR triggers offset markets that are thin and expensive. Title V triggers if total facility-wide emissions of any pollutant pass 100 tpy (or 50 / 25 tpy in nonattainment zones). The developer often selects a favorable jurisdiction or claims an emissions classification that does not survive scrutiny at the state agency.
Source. EPA Green Book NAAQS designations. State environmental agency permitting databases (see Part IV).
Counter-question. “What is the airshed designation for this parcel, the NSR threshold applicable at this location, the cumulative emissions from backup generation at full-load monthly test cycles, and the state environmental agency’s preliminary determination on the Plan Approval pathway? Is there a Title V trigger?”
12. “We will fund local schools and community programs.”
The truth. The headline number ($2 million over 10 years is a common form) is the package. It is presented in language that resembles a household budget. The number against the tax abatement is trivial. A $2 million package against $150 million of foregone state sales tax and $200 million of foregone county property tax over the abatement period is on the order of one half of one percent. The benefit package is also routinely tied to the abatement period and expires when the abatement does. After that, the developer has no reason to fund and no contractual obligation to.
Source. The developer’s own proffer documents. The county’s own auditor’s calculation if performed.
Counter-question. “What is the 30-year present value of the community benefit package, against the 30-year present value of the tax benefits the developer is receiving? Is the benefit package binding on the developer’s successors in interest, and does it survive the expiration of the abatement?”
13. “AI demand requires this facility now. Delay costs us the global race.”
The truth. AI demand projections are real but routinely inflated. Hyperscalers cancel and pause projects routinely. The 2025-2026 wave of hyperscaler capex revisions (Meta delays in the Midwest, Microsoft pulled-back leases in multiple geographies, Apple’s procurement pace on AI silicon, Stargate’s actual deployment behind its January 2025 announcement) all show the demand curve flexes by 30 to 50 percent in a single quarter on a single product release. The “urgency” framing is a tactic, not a constraint. The developer who needs the project approved this month cannot afford for the community to take three more months to read the documents.
Source. Public hyperscaler capex reporting. Vesper’s earlier coverage on Stargate’s actual deployment versus the announced numbers.
Counter-question. “What does this project look like if AI demand grows at half the currently projected rate rather than 100 percent? Is the developer willing to delay construction by 24 months while the community completes its review? If not, why is the timeline binding when the tenant has not been disclosed?”
SOUND
Counter-question. “Will the developer commit to ambient sound limits measured at the property line and at the nearest residential receptor, with continuous third-party metering, and remedies if exceeded? What is the maximum decibel level at 50 feet during full-load operation? Was the submitted sound study conducted under simulated full-load conditions with all generator load-test cycles running?”
The facility operates quietly. You will not notice it.”
The truth. Hyperscale data centers run 24/7 cooling at substantial decibel levels. Chiller plants, transformer hum, generator load-test cycles, and emergency backup systems combine to produce a low-frequency ambient that travels across property lines.
Documented complaints exist in Loudoun and Prince William Virginia,
Manassas Virginia,
Mesa Arizona,
Council Bluffs Iowa, and Chandler Arizona.
Sound studies submitted with applications are routinely conducted under non-representative conditions and exclude the full-load and full-test scenarios that produce the worst sound.
Source. Local journalism on Loudoun, Prince William, Manassas, Mesa, Council Bluffs, Chandler. ANSI S12.9 standards.
8. “Microsoft, Amazon, Google, or Meta has expressed strong interest in this site.”
The truth. Hyperscalers do not “express strong interest” through real estate developers at public meetings. Real hyperscaler interest produces a master agreement, an LOI with confidentiality, a queue request with the RTO, fiber commitments, and very specific technical requirements written into the design.
The developer is typically marketing a speculative site in the hope of attracting a hyperscaler later.
The probability of a specific named tenant materializing on a speculative site is not high.
The speculative campus that gets built without a tenant becomes the speculative campus that sits half-empty, the abatement is consumed, and the host has no leverage.
Dear @Erin Brockovich - around 200 of us marched earlier this month to hold the California Victims Compensation Board accountable for systemic illegally delays & rejection of benefits, ADA violations, due process violations, essentially RE-VICTIMIZING THOUSANDS of survivors of violence every year. Despite increases in funding, and federal funding, Cal VCB continues to pay out less victims. We are only 36 signatures away from 1000. Can you help a proud, autistic rape survivor protect others like him? Note the San Diego Mosque victims quality; please don’t let them be re-victimized.
Hold the city council accountable, talk to and email them and your local senate representative at your local meetings, keep records. ( if part if this is air pollution..Maybe buy your own EPA air pollution meter, like a PurpleAir meter that our family bought, and prove the source of the pollution with scientific proof of one layer on air pollution meters, if that’s a part of this ). I was able to do this in San Diego. Try looking online at the worldwide pollution meters on a website like PurpleAir.com. There might already be meters near the data centers or near your neighborhood).
It took me 9 years to prove a source of pollution near our home. After 18 years of our neighbors coming down with unusual life changing illnesses and deaths, it was obvious, but
Ohio EPA at epa.ohio.gov. Division of Air Pollution Control permits are searchable. Permit-to-Install and Permit-to-Operate are the two relevant categories.
Other states. Search “[state] air permit search” or “[state] DEQ public notice” or “[state] DEP air permit.” Every state has it. The interface varies. The data does not.
What to look for. Two things. First, whether a permit application exists for the project at all. Second, whose name is on it. The applicant identity is one of the most informative pieces of public data for understanding how the deal is actually structured.
A permit filed under the data center operator’s legal entity name signals a direct-build, where the developer or operator owns the on-site generation and the math runs through the developer’s own books.
A permit filed under a utility name signals a regulated-utility build, where the utility owns the plant and recovers the cost through its rate base. A permit filed under an independent power producer or a separate generation LLC signals a merchant or contracted-power structure, with a separate operator selling the output back to the data center under a power purchase agreement. Each structure produces a different answer to who pays what when the contract expires.
The second thing to look for is the nonattainment overlay. In counties designated as nonattainment for any criteria pollutant, the application triggers Nonattainment New Source Review (NNSR). NNSR requires emission offsets, a more demanding Lowest Achievable Emission Rate analysis instead of the looser Best Available Control Technology standard used in attainment areas, and longer agency review. Developers rarely disclose NNSR exposure in their hearing presentations because the offsets are expensive and sometimes unavailable in thin local markets, and the longer review changes the project’s economics and timeline.
The third use of the air permit is as a description of the equipment. The application narrates the generation technology in plain language: gas turbine make and model, generator class, peak power output, chiller plant capacity, cooling architecture. The community can read the cooling design off the air permit even when the developer’s hearing slide hedges on which architecture is “currently planned.” If the slide says “evaluating closed-loop” and the permit application describes evaporative cooling, the permit application is the binding statement of what is actually being built.
MATH
Georgia Board: Bottom Line
The Georgia Department of Audits stated explicitly in the 2023 review that the exemption’s cost-effectiveness was difficult to assess and that the foregone revenue substantially exceeded the documented economic benefit.
Virginia: Bottom Line
Virginia JLARC reached similar conclusions in 2024 on Virginia’s incentive program. Both auditors are nonpartisan state offices reporting to legislatures of mixed control. Both reports are public. Most county commissioners voting on data center abatements have not read either.
Ask For Detailed Answers
10. “We will be operational in 18 months.” Sometimes 24.
The truth. From conditional use approval to power-on, even on the most aggressive realistic timelines, hyperscale takes 36 to 60 months. The constraints are physical. Large power transformers ran 128 to 160 weeks of lead time through 2025 and have not normalized. Medium-voltage switchgear runs 40 to 60 weeks. Caterpillar G3520K reciprocating generators were booking into 2029 in early 2026. PJM cluster studies run 18 to 36 months and the deliverability assignment is the gate to construction. Interconnection upgrades on the utility side are not on the developer’s schedule. Environmental review can add 6 to 12 months.
Source. PJM interconnection queue documentation. ERCOT generator interconnection queue. EPRI and Wood Mackenzie reports on grid equipment lead times.
Counter-question. “Provide the developer’s critical path schedule, by phase, with the long-lead equipment delivery dates, the interconnection study completion date, the utility-side upgrade construction schedule, and the substation energization date. Does the developer hold signed purchase orders (not quotes) on the long-lead equipment?”
11. “The facility’s air quality impact is negligible. Backup generators are tested infrequently.”
The truth. In ozone nonattainment areas, backup diesel generators alone can push a facility over New Source Review thresholds. The 25 tons-per-year nitrogen oxides threshold applies in Severe ozone nonattainment areas: the Philadelphia 5-county area, the Houston-Galveston-Brazoria area, the Dallas-Fort Worth 10-county area, parts of the Northeast Corridor, the LA Basin. NSR triggers offset markets that are thin and expensive. Title V triggers if total facility-wide emissions of any pollutant pass 100 tpy (or 50 / 25 tpy in nonattainment zones). The developer often selects a favorable jurisdiction or claims an emissions classification that does not survive scrutiny at the state agency.
Source. EPA Green Book NAAQS designations. State environmental agency permitting databases (see Part IV).
Counter-question. “What is the airshed designation for this parcel, the NSR threshold applicable at this location, the cumulative emissions from backup generation at full-load monthly test cycles, and the state environmental agency’s preliminary determination on the Plan Approval pathway? Is there a Title V trigger?”
12. “We will fund local schools and community programs.”
The truth. The headline number ($2 million over 10 years is a common form) is the package. It is presented in language that resembles a household budget. The number against the tax abatement is trivial. A $2 million package against $150 million of foregone state sales tax and $200 million of foregone county property tax over the abatement period is on the order of one half of one percent. The benefit package is also routinely tied to the abatement period and expires when the abatement does. After that, the developer has no reason to fund and no contractual obligation to.
Source. The developer’s own proffer documents. The county’s own auditor’s calculation if performed.
Counter-question. “What is the 30-year present value of the community benefit package, against the 30-year present value of the tax benefits the developer is receiving? Is the benefit package binding on the developer’s successors in interest, and does it survive the expiration of the abatement?”
13. “AI demand requires this facility now. Delay costs us the global race.”
The truth. AI demand projections are real but routinely inflated. Hyperscalers cancel and pause projects routinely. The 2025-2026 wave of hyperscaler capex revisions (Meta delays in the Midwest, Microsoft pulled-back leases in multiple geographies, Apple’s procurement pace on AI silicon, Stargate’s actual deployment behind its January 2025 announcement) all show the demand curve flexes by 30 to 50 percent in a single quarter on a single product release. The “urgency” framing is a tactic, not a constraint. The developer who needs the project approved this month cannot afford for the community to take three more months to read the documents.
Source. Public hyperscaler capex reporting. Vesper’s earlier coverage on Stargate’s actual deployment versus the announced numbers.
Counter-question. “What does this project look like if AI demand grows at half the currently projected rate rather than 100 percent? Is the developer willing to delay construction by 24 months while the community completes its review? If not, why is the timeline binding when the tenant has not been disclosed?”
SOUND
Counter-question. “Will the developer commit to ambient sound limits measured at the property line and at the nearest residential receptor, with continuous third-party metering, and remedies if exceeded? What is the maximum decibel level at 50 feet during full-load operation? Was the submitted sound study conducted under simulated full-load conditions with all generator load-test cycles running?”
The facility operates quietly. You will not notice it.”
The truth. Hyperscale data centers run 24/7 cooling at substantial decibel levels. Chiller plants, transformer hum, generator load-test cycles, and emergency backup systems combine to produce a low-frequency ambient that travels across property lines.
Documented complaints exist in Loudoun and Prince William Virginia,
Manassas Virginia,
Mesa Arizona,
Council Bluffs Iowa, and Chandler Arizona.
Sound studies submitted with applications are routinely conducted under non-representative conditions and exclude the full-load and full-test scenarios that produce the worst sound.
Source. Local journalism on Loudoun, Prince William, Manassas, Mesa, Council Bluffs, Chandler. ANSI S12.9 standards.
8. “Microsoft, Amazon, Google, or Meta has expressed strong interest in this site.”
The truth. Hyperscalers do not “express strong interest” through real estate developers at public meetings. Real hyperscaler interest produces a master agreement, an LOI with confidentiality, a queue request with the RTO, fiber commitments, and very specific technical requirements written into the design.
The developer is typically marketing a speculative site in the hope of attracting a hyperscaler later.
The probability of a specific named tenant materializing on a speculative site is not high.
The speculative campus that gets built without a tenant becomes the speculative campus that sits half-empty, the abatement is consumed, and the host has no leverage.
Dear @Erin Brockovich - around 200 of us marched earlier this month to hold the California Victims Compensation Board accountable for systemic illegally delays & rejection of benefits, ADA violations, due process violations, essentially RE-VICTIMIZING THOUSANDS of survivors of violence every year. Despite increases in funding, and federal funding, Cal VCB continues to pay out less victims. We are only 36 signatures away from 1000. Can you help a proud, autistic rape survivor protect others like him? Note the San Diego Mosque victims quality; please don’t let them be re-victimized.
Link: https://www.change.org/p/state-federal-congressional-investigation-into-the-california-victims-compensation-board
Profit over safety. Don’t give up..
Hold the city council accountable, talk to and email them and your local senate representative at your local meetings, keep records. ( if part if this is air pollution..Maybe buy your own EPA air pollution meter, like a PurpleAir meter that our family bought, and prove the source of the pollution with scientific proof of one layer on air pollution meters, if that’s a part of this ). I was able to do this in San Diego. Try looking online at the worldwide pollution meters on a website like PurpleAir.com. There might already be meters near the data centers or near your neighborhood).
It took me 9 years to prove a source of pollution near our home. After 18 years of our neighbors coming down with unusual life changing illnesses and deaths, it was obvious, but