The Invisible Trillion · The Highway in the Bronx

Why the Numbers Lie

Before the legal machinery, one piece of technical sleight-of-hand, because it runs under the whole story — and once you see it, you can spot it in every claim that a facility is “within regulatory limits.”

Go back to Brittany Heights for a moment. The residents complained; the city measured; the meter reported that the data center was quieter than a normal conversation, somewhere around 50 decibels. The reading was accurate. It was also why nothing happened. The catch is that the meter and the body are measuring two different things, for three reasons worth carrying with you.

The first is that it never stops. A conversation lasts minutes. The hum runs every minute of every hour of every year, across an operating life now expected to stretch thirty to fifty years, and the body handles those two cases differently. A noise you know will end is one your nervous system can set aside; a noise that has run for eight years and will run for fifty more is one it never stops bracing against.

The World Health Organization recommends that nighttime outdoor noise stay under 40 decibels on average, because above that, studies tie it to high blood pressure, heart disease, broken sleep, and — in children — measurable drops in reading comprehension and memory. The European Environment Agency has called environmental noise the second-largest environmental cause of ill health in Europe, after air pollution. Data center noise routinely runs past that nighttime guideline, and after dark it stands out more, not less, as the world around it goes quiet.

The second is that it’s the wrong kind of sound to measure the usual way. The scale built into nearly every noise ordinance — A-weighting — was designed in the 1930s to gauge factory hearing damage, so it deliberately discounts low frequencies, which don’t harm hearing the way high ones do. Sensible for the 1930s problem. But a data center’s noise — thousands of cooling fans over the deeper rumble of chillers and pumps — lives mostly between 50 and 200 hertz, exactly where the A-weighted meter looks away.

So the official number understates what people actually hear. As one noise-control engineer put it in a 2025 trade publication, a low hum “contributes little to the dBA level, but causes complaints.” The most maddening part of the sound is the part the law doesn’t count.

The third is that it goes through walls. High-frequency sound bounces off windows and insulation; low-frequency sound passes through them, the way a neighbor’s bass reaches you when the melody doesn’t. That is why Donna Gallant’s triple-pane windows didn’t save her — they block the high notes the brochure emphasizes, not the low rumble she lives with. The same is true of most of what operators install under pressure: sound walls, acoustic louvers, rows of trees, all aimed at the high frequencies. The measures that work on the low end cost more and show up less often.

Hold those three, because the same move repeats through everything that follows: technical compliance with a measurement that misses the actual harm. Air emissions get checked at the fence line, on metrics that don’t capture what accumulates over a neighborhood. Water draws are measured against permits that ignore how long an aquifer takes to refill. Rate increases are modeled in ways that never ask who caused the new cost. Light mostly isn’t measured at all.

Nobody is lying. The regulators are using the tools they have, and the operators are following the rules as written — but the rules were written before the thing they regulate existed in this form. A noise ordinance built for backyard parties doesn’t have to be anyone’s deliberate gift. It just has to sit on the books, unchanged, while the facility is sited in perfect compliance with it.

All 19 chapters of The Invisible Trillion · Library · All courses