The Digital Deluge: Artificial intelligence’s hidden cost in water, watts, and carbon -
Artificial intelligence infrastructure is driving significant increases in water, energy, and carbon consumption, with a Cornell University study detailing the state-by-state environmental footprint. New Jersey is experiencing a boom in data center development, exemplified by projects from CoreWeave in Kenilworth and DataOne and Nebius in Vineland. Senator Bernie Sanders has proposed legislation to give the public a stake in AI companies and regulate their societal impact.
The rapid expansion of artificial intelligence infrastructure is incurring substantial environmental costs in terms of water, energy, and carbon emissions, a phenomenon dubbed the "digital deluge." A three-year study led by Fengqi You, a professor at Cornell University, and published in Nature Sustainability, provides the first state-by-state accounting of AI's environmental footprint, revealing that by 2030, AI data centers could add the equivalent of 5 million to 10 million cars to the nation's roadways in carbon emissions and consume enough water for 6 million to 10 million Americans annually. The study highlights that strategic data center siting in regions with lower water stress and cleaner energy grids, combined with improved cooling efficiency, could significantly reduce these impacts.
New Jersey is a hotspot for this development, with over 80 data centers operating or in progress. Specific projects include CoreWeave's $1.8 billion, 140-megawatt AI data center in Kenilworth and a hyperscale 300- to 400-megawatt facility in Vineland by DataOne and Nebius, which secured local tax incentives. Companies like OpenAI and Google are rapidly building AI data centers, contributing to the demand. The research suggests that even with grid decarbonization, total emissions could rise if AI demand outpaces the clean-energy transition, urging an acceleration of clean energy deployment alongside AI expansion.
Beyond environmental concerns, Senator Bernie Sanders introduced the American AI Sovereign Wealth Fund Act, proposing a one-time tax on large AI companies' stock to create a $7 trillion fund, giving the public a 50% ownership stake and potential annual dividends. This legislation aims to provide democratic oversight of AI development and mitigate potential negative societal impacts on jobs, privacy, and mental health. Experts like Mahmut Kandemir of Penn State warn that data centers could account for 20% of global electricity use by 2030-2035, emphasizing the urgency of sustainable development and regulation.