Engine certification
Tier 2
This is the federal certification level of the CAT 3516E engine itself. “Tier 4-equivalent” does not change that certification label.
Generators, fuel, emissions and sound
The fleet is large. Its categories and rules are specific.
Amazon’s application lists 766 critical and 16 support emergency generators. “Tier 4-equivalent controls” describes the controlled emissions performance proposed for the critical units—not the certification label of the underlying Tier 2 engines.
Generator breakdown
782 total units
| Type | Model | Count | Rating | Tier / controls | Role |
|---|---|---|---|---|---|
| Critical emergency generators | CAT 3516E | 766 | 2,750 kW | Tier 2 engine with Tier 4-equivalent controls | Critical data-center backup |
| Water/fire ancillary generators | CAT 3512C | 2 | 1,600 kW | Tier 2 | Water-cooling and fire emergency support |
| House emergency generators | CAT C18 or C27 | 11 | 750 kW | Tier 2 | House systems |
| Central administration generator | CAT C13 | 1 | 400 kW | Tier 3 | Administration building |
| Access-control generators | CAT C9 | 2 | 250 kW | Tier 3 | Access-control buildings |
What the control label means
Tier 2-certified engine plus proposed aftertreatment
The underlying 2,750 kW critical engines are Tier 2-certified. Amazon proposes adding selective catalytic reduction (SCR) and a catalyzed diesel particulate filter (cDPF) to meet what the application calls Tier 4-equivalent emission standards.
Proposed control train
SCR + cDPF
SCR targets nitrogen oxides. The catalyzed particulate filter targets particulate matter and also contributes to control of carbon monoxide and hydrocarbons.
Values used in Amazon’s calculations
Conservative control efficiencies in the submitted application
| Pollutant | Control efficiency used for potential-to-emit | Application context |
|---|---|---|
| Nitrogen oxides (NOx) | 91.7% | Lowest listed value among the three proposed systems. |
| Carbon monoxide (CO) | 69.8% | Conservative value selected for the calculations. |
| Volatile organic compounds (VOC) | 40.0% | Conservative value selected for the calculations. |
| Filterable particulate matter (PM) | 68.6% | Manufacturer options listed in the application range from 68.6% to 75.6%; the final system had not been selected. |
Why this matters: The submitted Hobart record does not establish a 97% particulate reduction. These are applicant calculations and proposed controls, not final IDEM findings. Verify on application p. 7
Scale and fuel
Capacity is not the same as annual use
Backup nameplate capacity
464 MW at SBN520 plus 1,655 MW at SBN750.
Gallons of listed storage capacity
Belly tanks plus 26 bulk tanks; not a forecast of annual burn.
Data-center buildings
Six at SBN520 and twenty at SBN750.
Proposed annual emission caps
Requested limits are not final permit conditions yet
These are Amazon’s requested combined caps for critical emergency generators CEG-1 through CEG-766. They become enforceable only if IDEM includes them in an issued permit.
Nitrogen oxides (NOx)
Combined cap proposed in the submitted application.
Carbon monoxide (CO)
Combined cap proposed in the submitted application.
Volatile organic compounds (VOC)
Combined cap proposed in the submitted application.
PM, PM10 and PM2.5
Combined cap proposed in the submitted application.
Sound evidence and local limits
A legal boundary standard exists; neighborhood levels still need measurement.
Adopted Hobart standard
65-decibel property-line maximum
Hobart Ordinance 2025-22 says data-center noise may not exceed 65 decibels at the applicant’s property line, except during emergencies caused by power outages. It also retains octave-band limits and a nuisance provision for irregular or intermittent noise.
Project agreement detail
Outage operation is an express exception.
The January 2026 community agreement separately states a 65 dBA project-boundary limit and lists exceptions including emergency repairs, public utility work and temporary backup generation during utility outages.
Sound evidence
The application includes manufacturer acoustical material.
Some later pages are scans or image-heavy equipment documents. They describe equipment-level assumptions; they are not a substitute for a final, independently verified neighborhood sound study at every home.
Responsible conclusion
A single site-wide decibel answer is not established.
Sound at a receiver depends on which units run, enclosure design, distance, barriers, terrain and background conditions. The application can inform review, but it does not justify claiming one universal decibel level “across from homes.”