Data Center Noise Lessons from Northern Virginia

Purpose of this report

This report gives a summary of the data center noise issue that has appeared in parts of Northern Virginia. The goal is to help county staff understand why some data centers create resident complaints even when they may meet a standard dBA noise limit, and how DCIP can avoid the same problem by designing for tonal and low-frequency sound before construction and then testing after the facility is operating.

The most important lesson is simple: most data centers are not causing public noise problems, but a small number have created serious resident concerns. The problem is usually not that a data center is generally loud like a factory or a highway. The problem is often a steady tone, hum, whine, rumble, or drone that runs day and night. Residents notice this type of sound because it stands out from the background, especially at night. A project can be below a standard dBA limit and still create a sound that people experience as annoying or intrusive.

A short lesson on acoustics: dBA, dBC, and tonal sound

Sound is measured in decibels, but not all decibel measurements describe the same thing. A-weighted decibels, written as dBA, are the most common way local noise ordinances measure sound. dBA is useful because it roughly follows the way the human ear hears many everyday sounds, especially speech and mid-frequency noise. For ordinary community noise, dBA is a reasonable starting point.

C-weighted decibels, written as dBC, are different. dBC gives more weight to lower-frequency sound. Low-frequency sound includes rumble, drone, and hum. This matters because large mechanical equipment, cooling fans, turbines, transformers, and exhaust systems can produce low-frequency sound that is not fully reflected in the dBA number. In other words, 55 dBA does not tell the whole story if the sound has a strong low-frequency component.

Tonal sound is another important idea. A tonal sound is not just a general whoosh of air. It is a sound with a clear pitch, such as a whine, buzz, hum, or drone. The human ear is very good at noticing tones. A steady tone can be more irritating than a louder but more random sound. That is why two sites with the same dBA level can feel very different to nearby residents.

For data centers, the right question is not only, “How many dBA is it?” The better question is: “Does the facility create a prominent tone or low-frequency hum at the property line or at the nearest homes?” Answering that question usually requires dBA, dBC, and octave-band or one-third-octave-band analysis. These methods show which frequency bands are causing the sound, instead of reporting only one total number.

Most Northern Virginia data centers do not appear to have noise complaints

This is an important point for county staff. Northern Virginia has the largest concentration of data centers in the world, yet the public examples of serious tonal or low-frequency noise issues are limited when compared with the total number of facilities. The Northern Virginia Regional Commission describes the region as having 250+ data centers. Governing reported in 2023 that there were nearly 300 data centers across Loudoun, Fairfax, and Prince William counties, with more in planning or under construction.

In this review, the strongest public examples of operational sound issues we identified and discussed in detail were Vantage, CloudHQ, and Compass/Vertiv. We also discussed other concerns, including Amazon/AWS Tanner Way in Prince William County, but the three examples below are the clearest for explaining the main lessons to county staff. Comparing three strong examples against roughly 250 to nearly 300 data centers is not a scientific complaint rate, because not every complaint is public and not every facility is the same size. But it does show the big picture: most data centers in Northern Virginia are not becoming major public noise cases.

Virginia’s own JLARC report reached the same general conclusion. It found that noise has been a problem for some data centers, but that a large majority do not generate noise complaints because of their location or design. That should be the practical goal for DCIP: not to deny that noise can be a problem, but to show the County how the project will be designed so it falls into the majority of facilities that do not create resident complaints.

How this has played out in Virginia

Northern Virginia has become the most visible U.S. example of data center growth near residential communities. The public record shows that not every data center creates noise complaints. The complaints appear to concentrate where a facility is close to homes and produces a continuous tonal or low-frequency sound from power generation, cooling equipment, fans, louvers, or airflow paths.

Virginia also shows that ordinary noise rules may not be enough. Prince William County materials have specifically discussed the limits of traditional A-weighted measurements for modern data center noise and have focused on low- frequency and tonal issues. Loudoun County best-practice discussions have also recognized the need to look at dBA and dBC, rather than only a single total sound reading.

Example 1: Vantage Data Centers in Sterling, Loudoun County

The Vantage example is the strongest warning for a power-backed data center. Public reporting has described the Vantage site on Glenn Drive in Sterling as a data center using eight natural-gas turbines while it awaits a Dominion electric connection. Residents complained about a constant whine, hum, buzzing, or tonal sound. Reports stated that the turbines were running day and night, which means residents did not receive nighttime relief.

The important lesson is that continuous on-site power generation can create a very different sound profile from a normal utility-fed data center. Backup generators that run only for testing or emergencies are one issue. Turbines running continuously are another. The sound must be controlled at the inlet, exhaust, enclosure, ventilation openings, vibration paths, and site layout. A simple wall may not solve the problem if the main source is low-frequency turbine exhaust or a strong tone.

For DCIP, the lesson is to design the power plant acoustically from the start. Any on-site generation should be reviewed before construction for dBA, dBC, and frequency-band performance. The design should include appropriate inlet and exhaust silencing, acoustic enclosure or equivalent treatment, vibration isolation, and post-startup testing. The County should not have to wait for residents to complain before the facility is evaluated.

Example 2: CloudHQ near Waxpool Road and Loudoun County Parkway

The CloudHQ example appears to be more of a cooling or mechanical-noise case. NBC Washington reported resident concerns about a loud, continuous droning noise near a CloudHQ data center in the Ashburn area. Residents compared the sound to a helicopter hovering over the neighborhood all day and night. This example is important because it shows that a noise problem can occur even without continuous on-site gas turbines.

For a conventional data center, the source may be cooling equipment, exterior fans, louvers, air-cooled chillers, rooftop equipment, or airflow paths. These sources can create a steady drone or hum if they are not selected, placed, isolated, or controlled correctly. Public reporting did not show a completed fix for CloudHQ, which makes it a useful example of why noise should be addressed during design, not after the community is already affected.

For DCIP, the lesson is that cooling and mechanical systems must be part of the acoustic design, not an afterthought. The facility should be modeled as a whole system, including power equipment, cooling equipment, transformers, and major airflow paths. The goal should be to avoid a prominent tone or low-frequency condition at the PUD boundary and at the nearest existing residential structure.

Example 3: Compass and Vertiv in Northern Virginia

The Compass and Vertiv case study is the best positive example because it shows a problem being identified and engineered out. Compass had a Northern Virginia data center where neighbors noticed a tonal sound even though the site was reportedly within a 55 dBA limit. The issue was not simply volume. The problem was a tone at lower frequencies that some nearby residents could perceive.

Vertiv and Compass studied the issue and traced it to an outdoor cooling unit support or fan-mount condition that affected airflow and created a tonal peak. The companies changed the design, including re-engineering fan mounts and making operational or software adjustments. Compass reported that the changes reduced sound pressure at lower frequencies and that follow-up sound studies found no tonality at neighborhood sites.

This example is very helpful for DCIP because it proves the point county staff need to understand: tonal noise is an engineering problem that can be measured, diagnosed, and reduced. It also shows why a simple dBA limit may not be enough. The sound can be legal on paper and still be objectionable if there is a distinct tone.

What these examples mean for DCIP

The Virginia examples point to a practical path forward. DCIP should not argue that data centers never create noise problems. Instead, DCIP should show that it understands the specific problem and is willing to design against it before operation. That approach is more credible and more protective of residents.

The County can protect residents while still allowing the project to proceed by focusing on objective, measurable standards. The PUD should require design-stage acoustic review for tonal and low-frequency sound; use dBA, dBC, and octave-band or one-third-octave-band analysis where appropriate; require reasonable design measures to prevent prominent tones; and require post-operation testing with a reasonable cure period if testing confirms a facility-caused tonal or low-frequency condition.

This approach avoids two extremes. It does not rely only on a general 55 dBA limit, which may miss low-frequency hum often better captured by dBC measurements or a pure tone. It also does not create a vague rule based only on subjective annoyance. Instead, it asks the right technical questions in a way that can be understood by county staff, residents, and the operator.

Conclusion

The Virginia experience shows that the issue is not just total loudness. The issue is whether a facility creates a prominent tone, hum, whine, rumble, or drone that carries to nearby homes. Most Northern Virginia data centers do not appear to create public noise complaints, and DCIP intends to be in that majority by addressing this risk during design, before construction, using qualified acoustic review and frequency-band analysis. After startup, DCIP will verify performance and correct any confirmed tonal or low-frequency condition attributable to the facility.

The Virginia cases are a constructive lesson. Vantage shows the risk of continuous on-site generation noise. CloudHQ shows the risk of cooling and mechanical drone. Compass and Vertiv show that tonal problems can be measured, diagnosed, and reduced when the source is identified and engineered correctly. DCIP can use those lessons to build a better project and give the County a stronger, more enforceable standard.

Selected source notes

Northern Virginia Regional Commission, Data Centers page, stating Northern Virginia has 250+ data centers. https://www.novaregion.org/1598/Data-Centers

Governing, “The Data Center Capital of the World Is in Virginia,” July 27, 2023, reporting nearly 300 data centers across Northern Virginia. https://www.governing.com/infrastructure/the-data-center-capital-of-the-world-is-in-virginia

JLARC, Data Centers in Virginia, 2024 report landing page and report. https://jlarc.virginia.gov/landing-2024-data-centers-in-virginia.asp

NBC Washington, Apr. 15, 2026, report on Vantage Data Centers in Sterling and gas turbine noise. https://www.nbcwashington.com/news/local/northern-virginia/amid-constant-data-center-noise-sterling-residents-also-worry-about-health-impact/4091393/

NBC Washington, Mar. 23-24, 2026, report on CloudHQ droning noise near Waxpool Road/Loudoun County Parkway. https://www.nbcwashington.com/news/local/northern-virginia/neighbors-raise-concerns-about-noisy-data-center-in-loudoun/4080249/

Compass Datacenters and Vertiv case study, “A Sound Solution,” regarding tonal sound mitigation at a Northern Virginia data center. https://www.compassdatacenters.com/compass-datacenters-and-vertiv-a-sound-solution/

Vertiv case study PDF on Compass/Vertiv sound attenuation work. https://www.vertiv.com/48ee77/globalassets/documents/case-studies/vertiv-compass-case-study-ch-00142.pdf

Prince William County, Data Center Ordinance Advisory Group noise ordinance update presentation, Mar. 18, 2025. https://www.pwcva.gov/assets/202503/DCOAG%20JMT%20Noise%20Ordinance%20Update%20Presentation%2020250318.pdf