Explainer

Noise

A data center is a building full of computers, and computers make heat. Getting the heat out is what makes the sound.

What it is, in plain words

Servers run all day and all night, so the cooling runs all day and all night too. Depending on the design that means banks of fans, chillers, or cooling towers on the roof and around the building. The sound people notice usually isn’t a bang. It’s a steady hum that doesn’t stop.

Backup power adds a second sound. Data centers are built to keep running when the grid doesn’t. Industry reliability tiers are defined partly by how much fuel is stored on site. A Tier III facility is built to hold 72 hours of backup fuel, and a Tier IV facility 96 hours.^[1] Those generators get exercised on a schedule. How often, and at what hour, is set by permit and by the operator, and we don’t have a Kentucky figure for it.

What the evidence actually shows

The most-reported measurements come from Northern Virginia, where the fight over data center noise has run longest. Reported, not verified. Facilities that drew noise complaints there measured between 40 and 59 decibels on nearby residential property. A well-designed modern facility typically runs 45 to 55 decibels at the property line.^[2] Residents in Prince William County, Virginia report levels routinely above 60 decibels.^[2] At the extreme end, some industrial sites have been reported at 96 decibels, around the clock.^[2]

Reported, not verified. Neighbors near data centers have reported headaches, vertigo, nausea, disturbed sleep, ear pain, elevated blood pressure, and a ringing sound. Residents in Arkansas and Texas are named specifically.^[2] These are complaints collected by reporters and researchers. They aren’t the output of a controlled health study, and we haven’t found one that measures health outcomes near data centers.

What’s contested or unknown

There’s a real technical problem underneath the argument. A lot of the sound is low-frequency, and some of it is below the range standard decibel meters capture well. That makes an ordinance hard to enforce. A facility can pass an A-weighted meter reading while neighbors still hear it inside their houses.^[2] That’s a measurement problem rather than a talking point, though it comes from the same single source as the numbers above.

Nothing has been built at either Carroll County site, so there is no Kentucky measurement to point at. Whether Virginia’s numbers transfer depends on the cooling equipment chosen and how far away the nearest house is. Terrain matters too, and so does whether there are berms or tree buffers in between. None of that is settled by a decibel figure from another state.

Questions worth asking at a hearing

• What sound level, in decibels, is predicted at the nearest property line at night with the facility at full load? Who produced that model, and will it be public?
• Which cooling system is planned, and does the noise model change if the design changes later?
• Is there a written noise limit in the agreement or permit? Measured where, at what hours, and by whom?
• Is low-frequency sound measured, or only A-weighted decibels?
• How often will backup generators run for testing, and at what time of day?
• Who pays for monitoring after the facility opens, and do residents get to see the readings?
• What happens if the limit is exceeded? Name the penalty and the person who enforces it.