The Persistent Hum of Hyperscale Data Centers
Understanding infrasound, low-frequency noise, and what communities should ask before approval
By Matt Legge, Founder, Meridian Group Strategies
When a hyperscale data center is proposed, public discussion often centers on electricity, water, land, and tax revenue. Noise deserves a place in that discussion, especially the persistent hum that can accompany operations running around the clock.
The term ILFN stands for infrasound and low-frequency noise. Understanding those terms helps residents and local officials ask better questions about a proposed facility, its equipment, and the sound that could reach neighboring homes.
What ILFN actually means
Infrasound refers to sound below 20 hertz, or Hz. Low-frequency noise describes the low-pitched portion of the sound spectrum, although assessment methods use different boundaries. EPA Victoria’s guidance, for example, evaluates low-frequency noise from 10 to 160 Hz. These categories overlap. [1][2]
Another relevant term is tonal noise. This describes sound concentrated at a particular frequency or narrow range of frequencies, such as a distinct hum or buzz. A tone can be low-frequency, but tonal noise and infrasound are not interchangeable.
That distinction matters. Hearing a persistent hum does not, by itself, establish that a facility is producing significant infrasound at someone’s home. That requires appropriate measurements.
Where data center noise comes from
Data centers depend on mechanical and electrical equipment. Cooling systems, HVAC equipment, backup generators, and, at some facilities, onsite power generation can contribute to the sound reaching neighboring properties.
Loudoun County, Virginia, identifies persistent tonal and low-frequency sound as challenges associated with data centers and their supporting infrastructure. The county also notes that tones can be more aggravating than broadband background sound at similar volumes. [3]
A project’s noise study therefore needs to identify the actual equipment and operating conditions. The label “hyperscale” alone cannot tell a resident what they will hear.
The relevant questions include which systems will operate continuously, how cooling equipment will perform under higher demand, and when generators or other power equipment will run.
Why one decibel number is not enough
A noise report may emphasize a single number expressed in dBA, or A-weighted decibels. A-weighting reduces the contribution of lower frequencies to the reported total. It is useful, but a single overall value can underrepresent the low-frequency or tonal character of a sound.
Loudoun County acknowledges this limitation and identifies additional tools, including C-weighted measurements and octave-band analysis, for evaluating a broader range of frequencies. [3]
In plain language, reviewers need to examine both the overall sound level and the frequencies making up that level.
A short daytime reading also leaves unanswered questions about nighttime conditions and changes in equipment operation. A meaningful assessment should explain when measurements were taken, where they were taken, and what the facility was doing at the time.
A qualified acoustical professional should select the methods and instruments, including equipment capable of assessing infrasound when that is part of the investigation.
What the research can and cannot tell us
An October 2026 research preprint reviewing hyperscale AI data center noise found that calibrated acoustic evidence remains limited. Cases with frequency detail commonly involved continuous tones, often around 70 to 140 Hz. Those frequencies are above the conventional infrasound boundary.
The authors caution that reported readings were collected under different conditions and cannot simply be combined into a representative community exposure level. The paper is a preprint and should be treated as emerging research. [4]
Health claims require similar care. A 2010 UK Health Protection Agency review found no clear evidence of a specific hazard from infrasound at levels commonly encountered in the environment, while acknowledging gaps in longer-term exposure research. That review was not a study of today’s hyperscale data center neighborhoods. [1]
The available evidence does not justify declaring every nearby household at risk from infrasound. It does support asking for better exposure measurements rather than making assumptions in either direction.
Residents’ reports of persistent sound deserve investigation. Determining its source, frequency, level, and duration is essential to understanding the problem and choosing an effective response.
What communities should request before approval
My recommendation is to make noise evaluation part of the approval process, with a clear plan for verification after operations begin.
Establish a baseline. Document existing daytime and nighttime sound at nearby noise-sensitive locations before construction. This gives reviewers a reference for evaluating future changes.
Evaluate the complete project. Study the proposed equipment and full campus buildout, including cooling demand, generator testing, and relevant onsite power operations.
Examine frequencies and tones. Assess low-frequency content and prominent tones alongside overall sound levels. Investigate infrasound with suitable instruments where warranted.
Assess affected properties. Explain what the model predicts at nearby homes and other sensitive locations, rather than relying exclusively on measurements near equipment.
Verify actual performance. After startup, test whether operations match the approved predictions, including representative nighttime and higher-load conditions.
Define responsibility for problems. Identify who investigates complaints, how findings are disclosed, and what corrective steps follow a confirmed problem.
These are recommended safeguards. Their legal form and enforceability need to be established through the applicable local approval process.
Florida should ask these questions early
For Florida communities reviewing hyperscale proposals, I believe the central question is straightforward:
What sound will reach neighboring properties, under what operating conditions, and how will those predictions be verified?
Residents should be able to review the assumptions behind a noise study. Officials should know whether that study addresses persistent tones and low frequencies. Developers should have clear expectations before choosing equipment and finalizing site designs.
A commitment to address complaints later leaves too much unresolved. Noise predictions, measurement methods, and responsibility for corrective action belong in the public discussion before approval.