AI data centres face scrutiny over environmental impact
As the rapid expansion of artificial intelligence data centres in the United States accelerates, experts warn that the environmental footprint—including high energy consumption, water usage, and air pollution—may outweigh the current societal benefits of generative AI.
The rapid expansion of artificial intelligence data centres in the United States has triggered significant scrutiny regarding their environmental footprint. Research suggests that power consumption for AI servers could reach between 165 and 326 terawatt-hours by 2028, contributing to air pollution that may result in billions in public health costs and thousands of premature deaths annually. Concerns also encompass the manufacturing of graphics processing units, which involves toxic chemicals and heavy metals, as well as the accumulation of electronic waste. While technology firms emphasise efficiency gains, experts argue that the environmental costs to local communities currently outweigh the societal benefits of generative AI.
The United States hosts more data centres than any other country, with tech companies racing to expand hyperscale facilities for AI. This growth has led to communities grappling with the prospect of having these large server warehouses as neighbours, raising concerns about utility bills and local pollution. The NAACP has sued xAI, now doing business as SpaceXAI, over air pollution from gas generators installed to power its massive data centres. The civil rights group has warned tech companies to be alert as it helps local groups across the nation mount campaigns against these environmental impacts.
Shaolei Ren, an associate professor at the University of California, Riverside, leads research into the impact data centres have on air quality and water resources for nearby communities. His study found that public health costs associated with the growing adoption of AI could reach more than $20 billion by 2028 and 1,300 premature deaths annually from air pollution by 2030. Ren advocates for a community-integrated data centre approach to prevent these facilities from harming nearby residents, noting that the current model often leaves local communities to bear the costs of progress.
Water scarcity is another critical issue, particularly in dry regions of the western United States. AI data centres require significant water for cooling systems, with usage spiking during hot weather. Ren estimates that US data centers could require up to 1,451 million gallons per day of new peak water capacity through 2030, a demand that would cost as much as $10 billion to meet. Small community water systems, often underfunded, may struggle to upgrade infrastructure without additional support, raising concerns about the burden placed on local resources.
The environmental toll extends beyond energy and water to the manufacturing and disposal of graphics processing units. Sophia Falk, a PhD candidate at Bonn University, has studied the material footprint of GPUs, finding that they are composed largely of heavy metals such as copper, iron, tin, and nickel. The production of these chips involves toxic chemicals, and the disposal of electronic waste poses further risks. Only about 22 percent of e-waste accumulating annually worldwide is collected and recycled, with much of the rest entering informal sectors where it can contaminate water sources and harm human health.
Tech companies highlight efficiency improvements, with Nvidia stating that its latest GPU architecture is significantly more efficient than previous versions. However, experts warn that increased efficiency can incentivise companies to burn through even more resources overall. Greenpeace East Asia is pushing Nvidia to clean up its supply chain by investing in renewable energy capacity where semiconductor manufacturing occurs. The organisation argues that the AI boom accelerates environmental phenomena in East Asia, where power grids are heavily reliant on fossil fuels.
As the debate over AI’s environmental impact continues, researchers and ethicists question whether the benefits of generative AI justify the costs. Catherine Flick, a professor of ethics and games technology at the University of Staffordshire, suggests that the industry needs to challenge the “bigger is better” mentality. She argues that the focus should be on determining what is usable and how to scope back investments to ensure they are worth the societal cost. The conversation highlights the need for greater transparency and accountability from tech companies as they navigate the ethical and environmental challenges of the AI era.
