Summary
- Residents near Microsoft’s Fairwater AI data centre in Wisconsin have filed a proposed class-action lawsuit alleging excessive noise, low-frequency hum, construction disturbance, and light pollution.
- Microsoft has said earlier tonal humming came from cooling fans operating at high speed during startup and that mitigations were put in place.
- The case is outside Europe, but the operational issues are familiar: cooling, acoustic treatment, light, and construction impacts are becoming consent risks for AI campuses.
Microsoft is facing a proposed class-action lawsuit over alleged noise and other local impacts from its Fairwater AI data centre in Mount Pleasant, Wisconsin, placing cooling equipment and community consent at the centre of a large AI infrastructure dispute.
The complaint, filed in federal court, has been brought by residents living near the site. The plaintiffs allege recurring and excessive noise, low-frequency hum, construction disturbance, dust, traffic, and light pollution. The claims have not been tested in court.
The allegations focus partly on diesel generators and HVAC equipment, including chillers, cooling towers, air-handling units, and condenser fans. The complaint also argues that some disturbance is a low-frequency hum that may not be fully captured by conventional A-weighted decibel measurements used in many municipal rules.
Microsoft has addressed sound complaints in its own community updates. In April, the company said residents north of the Fairwater 1 facility had reported a tonal humming sound after equipment began coming online. Its engineering team identified cooling fans operating at high speeds as the source.
In a June update, Microsoft said engineering teams and consultants had investigated the sound, conducted tests, and put mitigations in place. The company said several neighbours had confirmed what its monitoring showed: that the mitigations had resolved the issue. Additional sound-reduction components were scheduled for installation over the following months.
The Fairwater site is already operational at its first facility. Microsoft said in June that it had completed construction of the first Mount Pleasant building after bringing equipment online and conducting startup activities in April. A second adjacent facility is under construction, with completion scheduled for 2028.
Cooling leaves the fence line
The case sits outside the UK and Europe, but the infrastructure problem is not distant. Cooling design is becoming a public-facing issue for large AI data centres, especially where fan noise, tonal characteristics, night-time light, generator testing, construction hours, and traffic are experienced beyond the site boundary.
High-density AI facilities intensify the acoustic challenge. More heat has to be rejected, more plant may be visible or audible, and operating states can shift quickly during startup, commissioning, testing, or heavy load. Equipment that passes a decibel threshold can still create opposition if the sound has tonal or low-frequency characteristics that residents find intrusive.
European developers are already encountering a more difficult consent environment. Large projects in France, Germany, Ireland, the Netherlands, Switzerland, Scotland, and the UK are being scrutinised over power use, water, land, heat, noise, and local benefit. Acoustic modelling is moving from technical appendix to planning evidence.
The startup phase deserves particular attention. Commissioning can produce temporary conditions that differ from normal operation: fans at high speed, temporary plant, generator testing, construction lighting, dust, and heavy vehicle movements. Local opposition can harden during that phase, before a facility reaches its steady operating profile.
Consent becomes an operating condition
A technically resilient data centre can still carry project risk if the surrounding community experiences it as unmanaged industrial infrastructure. Legal claims, planning objections, mitigation costs, political scrutiny, and reputational pressure can all affect expansion programmes.
Developers need acoustic modelling that accounts for tonal and low-frequency sound, clear cooling equipment mitigation, disciplined construction management, lighting control, dust suppression, and a credible complaints route. Those measures are easier to design before equipment is ordered and buildings are phased than to retrofit after residents have organised.
The Fairwater claims remain allegations, and Microsoft says it acted to address earlier sound concerns. The wider lesson is operational. As AI campuses grow larger and closer to communities, airflow, vibration, fan control, and plant layout become part of the planning risk ledger.
