Are Data Centers Bad for Your Health?

Data centers pose no single dramatic health threat to nearby residents, but they are not benign neighbors either. The real risks are indirect and cumulative: degraded local air quality from backup diesel generators and fossil-fuel-heavy power grids, measurable increases in surrounding land temperatures, noise, and the potential for toxic gas release from on-site battery systems. None of these hazards are guaranteed to harm you, but communities living near large facilities have legitimate reasons to pay attention, especially as artificial intelligence drives an unprecedented wave of new construction.

Air Pollution From the Power Grid

The single largest health-relevant footprint of a data center is usually invisible: the electricity it consumes. A large facility can draw as much power as a small city, and where that power comes from determines whether nearby communities breathe dirtier air. In regions where the electrical grid still leans heavily on coal or natural gas, ramping up generation to feed a new data center means more pollutants leaving smokestacks. The pollutants that matter most for human health are fine particulate matter, sulfur dioxide, and nitrogen oxides, all of which are well-established drivers of cardiovascular and respiratory disease.¹

The connection is straightforward but easy to overlook. You do not see a smokestack on the data center roof. The emissions happen at a power plant miles away, and they get diluted across a region’s airshed. But epidemiologists have spent decades documenting how even modest, sustained increases in fine particulate matter raise rates of heart attacks, strokes, asthma exacerbations, and premature death. When a new facility adds hundreds of megawatts of demand to a grid that burns fossil fuels to meet it, the incremental pollution is real, even if no single resident can point to the data center as the sole cause of their symptoms.

This dynamic is not fixed, though. A data center plugged into a grid that runs mostly on wind, solar, hydroelectric, or nuclear power creates almost none of this burden. That is why the health implications of any particular facility depend enormously on geography. A campus in Quebec, where hydropower dominates, poses a fundamentally different air quality profile than one in West Texas, where natural gas still provides most generation capacity. Some operators sign long-term contracts to buy renewable energy, which can help, though how much it helps the local grid versus simply shifting clean power from one customer to another is still debated.

Diesel Backup Generators

Every major data center maintains rows of diesel generators as insurance against power outages. Uptime is the industry’s obsession: even a few minutes of downtime can cost millions of dollars and disrupt services for millions of people. The generators themselves sit idle most of the time, but they are not silent participants. Regulations require periodic testing, and each test run sends nitrogen oxides, particulate matter, and other combustion byproducts into the surrounding air. A recent assessment of data centers in Texas estimated that generator testing alone can release roughly 12 metric tons of nitrogen oxides per facility each year, compounding existing ozone problems in places like Houston and the Dallas-Fort Worth area.²

Twelve metric tons of nitrogen oxides from testing might sound modest compared to a busy highway interchange, but what makes it relevant is context. These generators are often clustered in areas that already struggle with ozone. Nitrogen oxides react with volatile organic compounds in sunlight to form ground-level ozone, which irritates airways and worsens asthma. When generators fire up during an actual grid emergency, which can last hours or even days, the emissions spike well beyond what routine testing produces. Climate-driven heat waves are making grid emergencies more common in exactly the regions where data center construction is booming, creating an uncomfortable feedback loop.

Construction-phase emissions deserve mention too. Building a hyperscale data center involves months of heavy earthmoving equipment, concrete trucks, and diesel-powered cranes. For communities adjacent to a construction site, the dust and exhaust during the build-out can be the most immediately noticeable impact, even if it is temporary.

The Data Heat Island Effect

Data centers convert virtually all of the electricity they consume into heat. Servers, storage drives, networking equipment: it all generates thermal energy that has to go somewhere. For most facilities, that means pushing enormous volumes of warm air into the atmosphere through rooftop cooling systems or cooling towers. When researchers examined satellite measurements of land surface temperature around AI hyperscale data centers worldwide, they found that the surrounding area warmed by about 2°C on average after a facility began operating.³

Two degrees of surface warming in the immediate vicinity might not sound like much, but it stacks on top of whatever urban heat island effect already exists. In cities and suburbs that are already several degrees warmer than surrounding rural land, an additional two-degree bump matters. Extreme heat kills more people in the United States than any other weather event, and its effects fall hardest on outdoor workers, elderly residents, young children, and people without reliable air conditioning. A localized hot spot does not need to be enormous to push vulnerable individuals closer to dangerous thresholds during a heat wave.

The warming pattern the researchers identified is distinct enough that they coined the term “data heat island effect” to describe it. Unlike the traditional urban heat island, which is driven by the accumulated thermal mass of roads, buildings, and vehicles, the data heat island is concentrated around a single point source. The temperature increase radiates outward from the facility and diminishes with distance, but it is persistent: the servers run around the clock, every day of the year, so the heat output never stops.

Noise

Cooling a data center requires fans, and lots of them. Large air-cooled facilities run thousands of industrial fans continuously, producing a low-frequency hum that can carry surprising distances, especially at night when background noise drops. Residents near data center campuses in Virginia’s “Data Center Alley” and in parts of the Netherlands and Ireland have filed noise complaints, reporting sleep disruption and chronic stress from the unrelenting sound.

Low-frequency noise is particularly tricky because it passes through walls and windows more easily than higher-pitched sound, and standard A-weighted decibel measurements undercount it. You might measure a reading at the property line that looks compliant with local noise ordinances, yet residents inside their homes still perceive a persistent drone. Chronic noise exposure is linked to elevated blood pressure, impaired sleep quality, and increased cortisol levels, and these effects do not require the noise to be loud enough to cause hearing damage. The threshold for sleep disruption is well below the level most jurisdictions regulate.

Some newer facilities mitigate this with sound barriers, enclosures around fan arrays, or a shift to liquid cooling systems that require fewer fans. But noise remains one of the most common and immediate quality-of-life complaints from data center neighbors, and it is often the issue that first draws a community’s attention to the facility’s broader impacts.

Battery Storage Fire Risks

As data centers integrate lithium-ion battery energy storage systems to smooth out power fluctuations and provide backup, they bring a fire risk that is still poorly captured by conventional safety frameworks. When lithium-ion batteries undergo thermal runaway, a cascading failure where one cell’s overheating triggers the next, they release hydrogen fluoride gas along with other toxic combustion products. Hydrogen fluoride is acutely dangerous: it irritates the lungs at low concentrations and can be lethal at higher ones.

A probabilistic risk assessment of a battery installation at a data center in Singapore modeled what would happen during a full thermal runaway event using different compartment designs and suppression strategies. The results were sobering: hydrogen fluoride concentrations exceeded the level considered immediately dangerous to life or health in every single simulated scenario, for both single-compartment and two-compartment designs. The study concluded that ventilation alone could not reduce the risk to acceptable levels for anyone present during such an event.⁴

For residents living near a data center with a large battery installation, the risk is not about day-to-day exposure. It is about what happens during a rare but catastrophic failure. Battery fires at industrial facilities have forced evacuations in several communities globally, and the toxic plume from a lithium-ion fire is not something you can simply shelter in place against. Fire departments are still adapting their protocols: lithium-ion fires behave differently from conventional fires, can reignite hours or days after appearing to be extinguished, and produce fumes that standard respiratory protection does not always filter effectively.

Electromagnetic Fields

Electromagnetic fields are a perennial worry for people living near any large electrical installation, and data centers are no exception. The concern typically centers on extremely low frequency fields produced by the transformers, high-voltage switchgear, and power distribution equipment that data centers require in abundance. This is an area where public anxiety runs well ahead of the measured risk.

Studies of high-voltage electrical infrastructure consistently find that electric and magnetic field levels, even at close range, fall below the reference safety levels set by international guidelines. Measurements at a substation in Bhutan found that even the highest recorded values, taken right at the transformer zone, remained within the safe range.⁵ A similar study of high-voltage transmission lines in Nigeria found that peak measurements stayed below 40 percent of the tolerable limits established by international guidelines.⁶

That said, the picture is not entirely settled. Researchers have flagged that investigation into long-term, low-level exposure has stagnated in recent years, partly because the consensus view is that these fields lack enough energy to damage cells directly. Yet some evidence suggests that chronic occupational exposure may still pose certain risks, even if the mechanism is not fully understood.⁷ For residents, as opposed to workers inside the facility, the exposure levels drop off sharply with distance. By the time you are across a street or behind a property-line buffer, the field strength from a data center’s electrical systems is typically negligible compared to the fields generated by your own home’s wiring and appliances.

Water Consumption and Local Supply

Large data centers are thirsty. Evaporative cooling towers, the most common method for rejecting heat from air-cooled facilities, consume millions of gallons of water per year. In water-stressed regions of the American Southwest, central India, and parts of northern China where data center construction is accelerating, this creates a direct competition with residential, agricultural, and ecological water needs.

The health connection is indirect but real. When a data center draws heavily on a municipal water supply, it can strain infrastructure that residents depend on for drinking water. In drought-prone areas, additional large-scale water consumption pushes the supply margin thinner, making water restrictions more likely and potentially driving up costs. Communities that depend on well water may experience drawdowns that lower their water table. None of this happens because the water a data center uses is itself hazardous. It happens because the resource is finite, and adding a very large new user to a limited supply has downstream consequences for everyone else.

Some operators have responded by investing in closed-loop cooling systems or shifting to direct liquid cooling, where a liquid coolant circulates through cold plates directly attached to server chips. This approach can dramatically reduce water consumption because the heat is transferred to the liquid without large-scale evaporation. Research into optimizing these systems is active: engineers are testing various coolant mixtures and manifold designs to handle the increasing thermal loads of AI chips.⁸ Whether these technologies are adopted fast enough to keep pace with the industry’s growth is an open question.

Where Data Centers Get Built and Who Bears the Burden

The health implications of living near a data center depend not just on the facility itself but on the community it lands in. Data center operators often seek cheap land, reliable power, and favorable tax incentives, which can steer them toward areas with less political power to resist. Rural communities and lower-income suburbs sometimes find themselves hosting massive campuses without having had much say in the decision. Researchers have called for empirical investigation into how data center expansion affects local populations, noting that existing evidence is thin and that communities near these facilities deserve better data on what they are actually being exposed to.¹

The environmental justice dimension is hard to ignore. A wealthy suburban community with strong zoning protections and organized civic groups is far more likely to block or condition a data center proposal than a rural county hungry for tax revenue and jobs. The jobs that data centers actually create tend to be few relative to the scale of the facility: once built, a hyperscale campus might employ a few dozen technicians. The community gets the tax base but also absorbs the noise, the heat, the water draw, and the generator emissions. Whether that tradeoff is fair depends on who is making it and whether they have full information about what “hosting a data center” actually means for daily life.

What the Research Still Does Not Cover

One striking feature of this entire topic is how little direct epidemiological evidence exists. There are no large cohort studies tracking health outcomes in populations living near data centers compared to matched control communities. The health concerns discussed in this article are largely inferred from what we know about the individual stressors: air pollution, noise, heat, and water stress each have deep epidemiological literatures, but almost none of that work has been applied specifically to data center neighborhoods. Researchers have explicitly identified this as a gap, calling the current situation one where a major new industrial sector is scaling at breakneck speed without the health surveillance infrastructure that typically accompanies it.¹

This does not mean the risks are imaginary. It means they are under-studied. The individual pathways are well understood: diesel exhaust causes respiratory harm, chronic noise disrupts sleep and raises cardiovascular risk, extreme heat kills vulnerable people, and hydrogen fluoride is acutely toxic. What we lack is the specific accounting that connects a data center to measurable changes in health outcomes in the surrounding population. That kind of research takes years, requires baseline health data collected before the facility arrives, and demands funding that has not materialized. Until it does, communities are left making decisions about data center proposals based on reasonable inference rather than facility-specific evidence, which is a frustrating position when the facilities in question are worth billions of dollars and backed by some of the most data-rich companies on the planet.