Datacenters Behaving Like Acoustic Weapons

Key scientific concepts and phenomena presented

• Infrasound: sound below ~20 Hz (below human hearing). The video treats infrasound as an environmental pollutant capable of physiological effects.
• Reported/linked health effects (from prior literature and interviews):
  • Elevated cortisol (stress / hypertension)
  • Vestibular effects: loss of balance, vertigo, dizziness, nausea
  • Vibroacoustic disease: pathological thickening of extracellular matrices in blood vessels, reducing brain blood flow (long-term claim cited from earlier literature)
  • High-frequency hearing loss, shortness of breath, anxiety, depression, increased cardiac workload
  • Eye irritation (cited as correlated in prior papers)
• Low-frequency man-made sources discussed:
  • Data centers (cooling/equipment)
  • Bitcoin-mining facilities
  • Oil & gas infrastructure (pumps, compressors, flares)
  • Fracking operations (induced seismicity and infrasonic emissions)

Measurement sites and sampling strategy

Five types of sampling locations were chosen:

1. Colossus data center (Memphis, TN)
2. Marathon Digital bitcoin-mining/data center facility (Hood County / Granbury area, Texas)
3. Permian Basin (West Texas / oil & gas fields) — used as a high industrial infrasonic baseline
4. Ambient/randomized samples across the U.S. (hotel rooms, houses, restaurants, wilderness)
5. Remote low-noise baseline (Death Valley — far from major infrastructure)

Observations:

• Infrasound amplitudes measured near the Memphis and Granbury data centers were substantially higher than ambient and even higher than Permian Basin samples.
• Comparisons used logarithmic scaling (because linear percent increases were unhelpful for these data).

Instrumentation and data processing methods

• Microphones with extended low-frequency response:
  • Flat down to ~9 Hz, usable to ~3 Hz
  • Require substantial wind protection
• Raspberry Shake 3D / seismic/vibration detectors:
  • Can record infrasonic and sub-3 Hz signals
  • Require more complex setup
• Small accelerometer / vibration sensors (industrial battery-monitoring sensors) used as portable infrasonic detectors
• Data processing steps:
  • Filter out >20 Hz to isolate infrasound
  • Time-scale (speed up) infrasonic recordings to render them audible for analysis/presentation
  • Convert vibration/seismograph output to Pascals (sound pressure) to treat as playable audio — an automated Python tool was written for this conversion (presenter intends to publish on GitHub)
• Comparative analysis methods:
  • Spectrograms, amplitude comparisons
  • Logarithmic regression for across-site comparisons

Double-blind human exposure experiment — design and results

Study setting and design:

• Location: synthesizer convention booth with an enclosed, insulated room
• Sessions: groups of five participants; sessions lasted ~3 minutes
• Cover story: participants were told the audio was “haunted” and asked to view a painting and listen
• Conditions:
  • 50% of sessions: infrasonic source ON (specialized subwoofer reproducing infrasound at ~25–30% amplitude of data-center field recordings)
  • 50% of sessions: control (infrasound OFF)
• Blinding: participants and on-site researchers were intended to be unaware of condition (double-blind)
• Survey: after exposure participants rated various symptoms/feelings on a 0–10 scale
• Sample: >100 participants initially; final clean double-blind dataset = 74 subjects (data cleaning excluded incomplete/suspicious responses)

Main effects observed (infrasound vs control)

• Increased likelihoods / mean differences (reported percent changes and some means):
  • Tingling / restless legs: +25%
  • Pain: reported only by a very small number, exclusively in the infrasound group
  • Tiredness: +12%
  • Nausea: +33% (means small: infrasound mean ≈ 1.2 vs control ≈ 0.9 on 0–10 scale)
  • Dizziness: +150% — a strong effect (means: infrasound avg ≈ 3.0 vs control ≈ 1.2)
  • Chills: +20%
  • Irritability / depression: +33%
  • Eye irritation: ~3× more likely (noted as requiring larger sample)
  • Lethargy: +80% (presenter questioned/discounted this metric as ambiguous)
  • Anxiety: +55%
  • Sadness: ~2× more likely
  • Discomfort: ~3× more likely (most striking result: mean discomfort control ≈ 1.2 vs infrasound ≈ 4.8)
• Decreases:
  • “Creeped out”: −11%
  • Feeling “spiritual”: −43% (means: control ≈ 4.2 vs infrasound ≈ 2.4)

Presenter caveats about the experiment:

Many absolute mean values were low (e.g., nausea mean ~1), so percent changes can overstate clinical importance. Dizziness and discomfort were highlighted as the most robust/meaningful outcomes from this pilot.

Environmental / industrial context and associated pollutants

• Colossus data center (Memphis)
  • Heavy electricity use (claimed large fraction of city capacity; expansions planned)
  • Use of methane gas turbines to offset grid draw — emissions of NOx, CO, SOx, VOCs (respiratory and cardiovascular impacts)
  • Large water usage (claimed ~1 million gallons/day; much returned as steam)
• Marathon Digital bitcoin facility (Hood County)
  • Widely reported loud audible noise and infrasound; local complaints and civil suits; reported effects on nearby animals/plants in local reports
• Permian Basin
  • Heavy flaring, venting, fracking
  • Reported induced seismicity (thousands of small earthquakes per year; many of the strongest recent TX quakes in gas regions)

Methodological notes and reproducibility

• Three capture approaches described (extended-mic, Raspberry Shake seismometer, small accelerometer), each with trade-offs in frequency range and practicality
• A Python pipeline was developed to convert vibration data to Pascals and to process many files; presenter intends to share code on GitHub
• Emphasis on need for:
  • Repeatability
  • Larger sample sizes
  • Frequency-specific testing
  • Formalized peer review and funding for rigorous follow-up studies

Recommendations, legal & policy implications

• Advocate treating infrasound monitoring/regulation similarly to air and water quality (set thresholds, require monitoring)
• Community action recommendation:
  • Install seismic/infrasound monitors (e.g., Raspberry Shake) to log activity and build evidence for complaints or litigation
  • Hard infrasonic logs could be used by class-action attorneys and regulators in lawsuits/regulatory actions, especially where official responses assert the noise predated construction

Notes on accuracy and caveats

• Subtitles used by the presenter are auto-generated and contain transcription errors (e.g., “viro acoustic” likely intended as “vibroacoustic” or “vibroacoustic disease”; some place-names and company names may be misspelled)
• The video mixes field recordings, anecdotal testimony, a pilot double-blind experiment, and policy commentary
• The human experiment is a pilot/demonstration (n = 74 clean double-blind), not a large clinical trial; the presenter explicitly calls for more rigorous, reproducible research

Researchers, people, institutions, products, and sources featured

• Individuals and groups:
  • Cheryl Shaden — resident interviewed living across from the bitcoin mine
  • Residents and litigants near the Granbury/Hood County crypto-mine (multiple local people; one family moved because of a child’s seizures)
  • Local officials referenced (mayor of Glenn Rose; county commissioner — names not always given)
• Organizations, places, and products:
  • Colossus data center (Memphis, TN)
  • Marathon Digital — bitcoin-mining/data center facility (Hood County, TX)
  • Permian Basin (West Texas / oil & gas industry)
  • Raspberry Shake (Raspberry Shake 3D systems)
  • Small accelerometer / battery-monitoring vibration sensors
  • “Ghost in the Machine” paper (prior paper referenced linking infrasound and eye irritation / other effects; exact citation not provided)
  • Synthesizer convention / organizers (appears in subtitles; exact name may be mistranscribed)
  • GitHub / Python tool (presenter intends to publish conversion/processing code)
• Note: a fictional or narrative LLM (“Grock” / Colossus references) and mentions of Elon Musk appear in the video narrative but are not scientific sources

Possible follow-ups (as presented)

• Extract the core quantitative results into a concise table (means and % differences) from the reported pilot experiment
• Provide suggested next steps for a rigorous follow-up study (sample size calculation, controlled exposure parameters, measurement instrumentation and calibration, clinical endpoints)