Are AI Data Centers a Health Concern?

With the public release of ChatGPT in late 2022, artificial intelligence (AI) has been thrust into the mainstream — incorporated into various industries as well as our personal lives. In healthcare, AI scribes are prominent, and peers use OpenEvidence as part of their clinical practice. While AI has its benefits, we cannot deny the immense health and environmental costs, which are only growing as more hyper-scale data centers (which can span more than 17 football fields) are rapidly developed.
Communities nationwide are fighting this. With moratoriums and calls for scaling back on development, we must take advantage of this opportunity to ensure the creation of safe policies and regulations. Data centers should be the size necessary to run basic functions but no larger; they must be sustainably built, and include transparency and community input throughout all stages of development; and they cannot perpetuate the harms and injustices that many communities have historically faced.
Energy and Pollution
Data centers require large amounts of energy to run AI processes. Unfortunately, in part because of their rapid development and need for energy, we have postponed closing fossil fuel generators that were slated to shutter in 2025.
When fossil fuels are burned, they emit pollutants that are linked to poor health outcomes such as premature death, lung diseases, cognitive decline, and heart attacks. An analysis of a Virginia data center found its air emissions could cost between $53 million and $99 million annually in health-related impacts. Another study found that AI’s contribution to air pollution could result in 1,300 premature asthma-related deaths in 2030.
Meanwhile, short-term exposure to PM_2.5 was attributed to 1 million premature deaths from 2000-2019, which is concerning given that these data centers are continuously spewing air pollutants. Furthermore, fence-line communities — predominantly Black, Latino, and low-income communities — already face disproportionate health impacts due to living near polluting infrastructures, and in some places are disproportionately sited to have more data centers developed near them compared to other communities.
Water Concerns
Data centers use copious amounts of potable water, not just to cool their servers, but also to generate the electricity that powers the centers. Many data centers are built in drought-prone areas, stressing communities already facing water scarcity, and increasing the risk of dehydration, waterborne illnesses such as dengue, and related mental health consequences.

Electricity generation drives much of data centers’ water usage. Between Texas, California, Georgia, Ohio, Illinois, Arizona, and Virginia, 3.4 trillion gallons of freshwater were used annually for data center electricity, which is about 12 times the water used to power data centers in Los Angeles, Washington, DC, and Phoenix combined.
Communities are also concerned about water pollution. For example, per- and polyfluoroalkyl substances (PFAS or “forever chemicals”) may leak and be released into soil and groundwater systems. PFAS are difficult to remove and have been linked to birth defects, cancer, and heart and brain diseases. Additionally, water used as a coolant can be released back into water supplies hotter than it was before, disrupting aquatic ecosystems and biological processes important to the functioning of our planet.
Noise Pollution
Diesel generators, cooling towers, and the machinery in data centers emit persistent noises that have caused distress to communities. The World Health Organization states that noise pollution is linked to increased mental health issues like heart attacks, strokes, or poor academic performance in children, and infrasound (waves that can be felt but not heard, and aren’t easily blocked by windows or buildings) are linked to anxiety, headaches, stress, vertigo, and insomnia.
Some parts of data centers, like cooling towers, can reach up to 70 decibels within 400 feet, and industrial-sized diesel generators can reach 105 decibels. Data centers continuously emit these noises and sound waves. And we anticipate that vulnerable neighborhoods already facing noise exposure inequities from transportation or aviation will continue to bear this burden with further development of data centers.
Though not as discussed as much, data centers should also take caution with how they are built, lest they contribute to light pollution, which impacts our circadian rhythm, which disrupts sleep, and is linked to impaired attention and memory, dementia, obesity, hypertension, and mood disorders like depression.
What Can We Do?
Health professionals can play an important role in responsible AI usage and expansion to prevent the perpetuation of injustices that put communities at further environmental and health risks. There are actions we can take now:

• Ask your institution if there is an AI governance board. If so, inquire how they’re deciding which AI tools to add to our workflows and if they’re obtaining user feedback on impact and efficiency. Which tools are extraneous and can be retired so as to decrease unnecessary AI usage?
• Discuss with the IT team if they’re working with AI vendors that focus on sustainability. How are we holding them accountable and ensuring they’re tracking carbon emissions?
• In your personal life, determine which AI tools you don’t need. One AI search query can use five times the amount of electricity compared to a normal web search query or 16 oz of fresh water. You probably don’t need to use AI for every query.
• Discuss environmental and health harms with colleagues or communities. Our voices are important, and we must advocate for our communities and patients.
• Encourage your friends, family, and patients to vote! See where candidates align on environmental issues, including AI.

While not an exhaustive list, we can take personal and community actions to advocate for responsible AI and data center development and fulfill the promise we made to do no harm.

The views of the author do not necessarily reflect those of the Medical Society Consortium on Climate and Health or its members.