Are Data Centers Making Their Neighbors Sick?
What the evidence shows about noise, sleep, and “dirty electricity”
Environmental Health
Thomas Lee Abshier, ND — 24 October 2026
My position
I stand with the neighbors. People living beside large AI data centers are reporting a consistent cluster of symptoms: broken sleep, headaches, a constant sense of being on edge, ear pressure, nausea, dizziness, and the feeling that their home is no longer a refuge. I take these reports seriously, and I believe the best-supported cause is the one easiest to overlook because it doesn’t register well on ordinary meters: constant, low-frequency noise and vibration, running day and night, with the sleep loss and stress physiology that follow.
A second proposed cause has drawn much attention: “dirty electricity,” the high-frequency distortion that data centers inject onto the power lines serving nearby homes. That the distortion exists near data centers is supported by real data. No evidence shows it harms human health. The question is legitimate and worth answering, and below I explain both why the evidence is weak and why the concern remains plausible.
This article was prompted by a widely read September 2026 essay (“How the Law of Displacement Can Solve the Data Center Crisis,” by A Midwestern Doctor) that argues the symptoms near data centers overlap with those reported by electrically sensitive people and attributes part of the problem to dirty electricity. I agree with much of the essay’s diagnosis and its proposed remedy (move the large facilities to remote sites with their own power). My task here is to sort the claims by the strength of their evidence, so that patients and communities can act on what is solid.
1. What neighbors are reporting
The pattern described by residents near data centers in Virginia, Ohio, Texas, and elsewhere is remarkably consistent:
- A constant hum, whine, or “vacuum cleaner left on” sound, often more noticeable indoors and at night
- Vibration felt in walls, windows, floors, or the chest
- Difficulty falling asleep, waking at night, waking early
- Headaches and a sense of pressure in the head or ears
- Irritability, anxiety, and fatigue the next day
- In some, nausea, dizziness, or worsening tinnitus
- Diesel odors during generator testing; worsened breathing in some
These are not trivial complaints. Chronically broken sleep isn’t a nuisance; it is a physiological injury that raises blood pressure, disrupts glucose regulation, impairs immunity, and worsens mood and pain.
2. The strongest explanation: low-frequency noise and sleep
Data center cooling equipment (large chillers, slow-turning fans, compressor arrays, transformers) produces broadband noise with a strong low-frequency component. Three facts make this kind of noise particularly harmful:
It is under-measured. Most noise ordinances use the A-weighted decibel scale, dB(A), which mimics the ear’s sensitivity at moderate volumes and therefore heavily discounts low frequencies. A low tonal hum can contribute little to the dB(A) reading while still causing complaints. A facility can pass its noise permit and still keep a family awake.
It penetrates buildings. Low-frequency sound passes through walls and windows far more readily than higher-pitched sound. Closing the windows helps less than people expect, which is why many residents report the hum seems louder inside.
It is constant. The World Health Organization’s environmental noise guidelines identify night-time noise as particularly harmful because sleep is when the body performs cardiovascular and neurological recovery. A sound that never stops denies the nervous system the chance to stand down.
Systematic reviews of low-frequency noise find associations with annoyance, sleep problems, concentration difficulties, and headaches. Evidence for chronic disease outcomes specifically from low-frequency noise is thinner, and claims about infrasound (sound below the range of hearing) at environmental levels remain contested. But the core phenomenon, chronic, inescapable low-frequency noise producing poor sleep, headaches, and stress physiology, is well established in general noise research and matches what data center neighbors describe. Diesel exhaust from generator testing adds a further, well-understood respiratory burden.
My clinical judgment is that this is where most of the harm is coming from, and it is the part we can measure and stop today.
3. Dirty electricity: what the data actually show
The electrical finding is real
In December 2024, Bloomberg News analyzed about 770,000 home power-quality sensors made by Whisker Labs and compared their readings with data center locations. More than half of the homes with the worst distortion were within 20 miles of significant data center activity, and more than three-quarters of the worst readings were within 50 miles. In Loudoun County, Virginia (“data center alley”), the share of sensors exceeding the 8 percent total harmonic distortion threshold was roughly four times the national average. The pattern held in rural areas as well as cities, and the worst distortion tended to occur at night, when data centers make up a larger share of demand.
This is consistent with basic electrical engineering: data centers draw power through vast numbers of switching power supplies, which distort the voltage waveform on the lines they share. Industry engineering literature acknowledges that these distortions can extend to other customers on the same infrastructure. The utilities serving the hardest-hit regions (Dominion in Virginia, ComEd in Illinois) disputed the findings, but the geographic correlation across many independent sensors is difficult to dismiss. I accept that data centers are degrading power quality in surrounding neighborhoods. At minimum, that harms appliances and motors and increases electrical stress on the grid.
The health link is not established
The leap from distorted power to human illness is a separate question, and here the evidence is weak.
First, Bloomberg data and the dirty-electricity health hypothesis do not measure the same thing. Bloomberg’s sensors measure voltage harmonics of the 60-cycle current, generally up to a few thousand cycles per second. The dirty electricity health literature focuses on higher-frequency voltage transients, roughly 4,000 to 150,000 cycles per second (the range Graham-Stetzer filters target). The Bloomberg analysis did not measure those higher frequencies. The two may well rise together near data centers, but no one has shown that they do.
Second, the direct health evidence for dirty electricity is thin. A 2016 systematic review in Frontiers in Public Health (de Vocht and Olsen) examined the peer-reviewed literature and found that the evidence rested mainly on case reports, self-reported improvements in unblinded filter installations, and ecological associations; it found no basis for the commonly used safety threshold and concluded that the evidence did not withstand scientific scrutiny.
Third, the broader research on electromagnetic hypersensitivity points in the same direction. A systematic review of 46 blinded or double-blinded provocation studies involving more than 1,100 self-identified electrosensitive volunteers found no robust evidence that people could detect fields better than chance or that symptoms tracked actual exposure. The same studies did find that expecting exposure reliably produced symptoms. People suffered genuinely, but the trigger appeared to be the belief, not the field.
That last point deserves care, not dismissal. Nocebo effects are real physiological responses, not imagination; they produce real headaches and real insomnia. But they also imply something practical and important for communities: if residents are told that invisible electricity is making them sick, more of them will become sick, and they will mislocate the cause. The noise, which can be measured and mitigated, may go unaddressed while people buy shielding that does nothing for a 40-hertz hum coming through the bedroom wall.
4. Why the concern remains plausible
That said, I don’t regard the dirty electricity question as closed, for four reasons.
No one has measured the right thing in the right place. Utilities monitor harmonics up to about 2–3 kilohertz. Data centers are known to emit in the higher “supraharmonic” band (2–150 kilohertz) as well, but this band is almost never monitored on distribution grids, and it has never been measured systematically in homes near data centers. An unmeasured exposure cannot be declared harmless.
The existing negative studies tested something different. The provocation studies mostly tested short exposures to mobile-phone radio signals or simple fields in a laboratory, over minutes or hours. Data center neighbors live with continuous, low-level, night-time exposure for months. Short-term blinded tests are good at answering whether someone can feel a field now; they are poor at answering whether chronic exposure disturbs sleep architecture or autonomic tone.
Related fields carry recognized signals. The extremely low-frequency magnetic fields from power lines are classified by the International Agency for Research on Cancer as possibly carcinogenic, based on a repeatedly observed association with childhood leukemia that remains unexplained. The new transmission lines being built to serve data centers will increase such exposure for some households. A possible-carcinogen classification is not proof of harm, but it is a reason for prudence in siting.
The incentives run against looking. The parties with the instruments, data, and money to measure power quality near data centers, the utilities and the operators, have every reason not to find a problem. The utilities responded to Bloomberg by disputing the data rather than releasing their own. When the only people who can settle a question have a financial interest in keeping it unsettled, a reasonable person keeps the question open.
5. What would settle it
The question is answerable, and the cost of answering it is small compared with the money being spent on data centers:
- Independent measurement. Monitor homes at varying distances from data centers across the full spectrum, including the 2–150 kHz band and magnetic fields, at night, over months.
- Measure the noise properly. Use C-weighted and one-third-octave measurements down to 10–20 Hz inside bedrooms at night, alongside the electrical measurements, so we can separate the two exposures.
- A blinded home trial. Install active filters in some homes and identical-looking dummy units in others, with neither residents nor investigators knowing which is which, and track sleep objectively (wearable actigraphy or home sleep recordings) and symptoms. If dirty electricity is harming sleep, real filters will outperform dummies. If they do not, families can stop spending money on them and focus on the noise.
6. Practical advice for patients near data centers
- Protect sleep first. Sleep in the room farthest from the facility. Low-frequency sound is poorly blocked by ordinary earplugs; mass (heavy curtains, sealed windows, an interior room) helps more than foam. A fan or broadband sound machine can mask the hum for some people.
- Document the noise correctly. If you are filing a complaint, insist on low-frequency measurements, not just dB(A). A night-time recording with a C-weighted reading is far more persuasive than a daytime dB(A) number.
- Address generator exhaust. Note generator test schedules; keep windows closed and run a HEPA/carbon air purifier during tests.
- If you want to test dirty electricity filters, test them honestly. Have a family member install either working filters or unplugged look-alikes without telling you which; keep a sleep and symptom diary for two weeks, then switch. If you cannot tell the difference blinded, the filters are not your answer.
- Do not ignore ordinary medicine. Headache, dizziness, and insomnia have many causes. A new or worsening symptom deserves a proper evaluation, not only an environmental explanation.
- Organize locally. The most effective protections are written into permits before construction: enforceable night-time low-frequency noise limits measured at the nearest home, generator testing restricted to daytime, setbacks, and requirements that large facilities generate their own power off the residential grid.
Conclusion
Data centers placed among homes harm their neighbors, and the clearest mechanism is constant low-frequency noise that disrupts sleep and keeps the nervous system on alert. Data centers also measurably degrade electricity quality in surrounding neighborhoods. Whether that electrical distortion harms human health remains unproven; existing evidence leans against it, but no one has measured exposure in the right homes over the right length of time, and those with the means to do so have reasons not to.
The remedy for both concerns is the same: move gigawatt-scale computing away from where people live and sleep, onto sites with their own power, off the residential grid. That removes the noise, distortion, water draw, and higher utility bills at once, without forcing us to settle every scientific dispute first.
Sources
- A Midwestern Doctor, “How the Law of Displacement Can Solve the Data Center Crisis,” The Forgotten Side of Medicine (Substack), September 6, 2026. https://www.midwesterndoctor.com/p/how-the-law-of-displacement-can-solve
- Bloomberg, “AI Needs So Much Power, It’s Making Yours Worse,” December 2024: https://www.bloomberg.com/graphics/2024-ai-power-home-appliances
- CleanTechnica, “Demand for Power at Data Centers is Disrupting Harmonics in Nearby Electrical Grids” (utility responses): https://cleantechnica.com/2024/12/30/demand-for-power-at-data-centers-is-disrupting-harmonics-in-nearby-electrical-grids/
- de Vocht F, Olsen RG. “Systematic Review of the Exposure Assessment and Epidemiology of High-Frequency Voltage Transients.” Front Public Health 2016;4:52: https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2016.00052/full
- Rubin GJ, Nieto-Hernandez R, Wessely S. “Idiopathic environmental intolerance attributed to electromagnetic fields: an updated systematic review of provocation studies.” Bioelectromagnetics 2010;31(1):1-11: https://pubmed.ncbi.nlm.nih.gov/19681059/
- Rubin GJ, Das Munshi J, Wessely S. “Electromagnetic hypersensitivity: a systematic review of provocation studies.” Psychosom Med 2005;67(2):224-32: https://pubmed.ncbi.nlm.nih.gov/15784787/
- Schmiedchen K, Driessen S, Oftedal G. “Methodological limitations in experimental studies on symptom development in individuals with IEI-EMF.” Environ Health 2019: https://pmc.ncbi.nlm.nih.gov/articles/PMC6805477
- World Health Organization, Environmental Noise Guidelines for the European Region, 2018.
- Irish Times letters, “Noisy data centres are far more than an annoyance,” September 2026: https://www.irishtimes.com/opinion/letters/2026/09/07/noisy-data-centres-are-far-more-than-an-annoyance/