Who’s afraid of the big, bad GPU?

The rapid expansion of generative artificial intelligence has triggered a massive global demand for graphics processing units (GPUs), bringing their severe environmental and social impacts into sharp focus.

Originally developed for video games, these specialized processors are now central to powering massive data centers, primarily across the United States. This surge in infrastructure development has raised alarms among local communities and researchers regarding the enormous quantities of electricity and water required to cool and run these facilities. High-profile civil rights organizations and environmental advocates have begun legal and public campaigns against tech firms, citing rising utility bills, local air pollution from backup generators, and severe water strain in drought-prone areas during peak cooling periods.

Beyond the immediate resource consumption of data centers, the hardware supply chain itself poses significant ecological hazards. Manufacturing GPUs relies on intensive mining of heavy metals like copper, which risks contaminating water sources with acidic runoff, while semiconductor fabrication facilities utilize toxic chemicals that historically polluted surrounding communities. Researchers studying these lifecycles warn that the rapid obsolescence of AI hardware will generate hundreds of thousands of tons of electronic waste by the end of the decade, much of which may end up in unregulated recycling markets in developing countries. While major chipmakers like Nvidia, Intel, and TSMC highlight recent efficiency gains and sustainability commitments, critics argue that the tech industry must shift away from its relentless drive for scale to genuinely address these systemic ecological costs.