AI Data Centers’ Energy and Water Consumption
Develop policy solutions that mitigate the environmental and societal impacts of generative AI infrastructure.
Project Leads
Marc Conte, Associate Professor Department of Economics
Kimberly Diamond, Adjunct Professor of Law
Project Description
Generative AI comprises up to hundreds of thousands of graphics processing units (“GPUs”). GPUs conduct computations almost continuously, generating substantial heat. To prevent GPU heat damage, AI data centers supporting generative AI (“ADCs”) require cooling systems, and options include open systems that consume massive quantities of water or closed systems that demand less water but use more energy. ADCs consume water in parallel with their energy consumption and cooling needs, enough to fill over 3,000 Olympic-size pools in certain cases. Colossal demand for generative AI user queries, driven by artificially-low prices not evidencing electricity’s and water’s true costs, has required construction of enormous, Manhattan-sized ADCs.
ADCs’ scale of accelerating energy demand and astronomical water consumption, coupled with unpriced impacts on human well-being via CO2 contributions to anthropogenic climate change, is highly problematic; causes adverse impacts to the environment, wildlife, Indigenous nations, and humanity generally; and poses an existential threat if not addressed expeditiously through immediate policy action (the “Issue”).
Due to a general lack of localized legal guardrails sufficient to address the Issue, divergent legal theories governing water rights nationally, and judges generally applying legal balancing tests myopically when weighing the economic utility of actors’ conduct against that conduct’s environmental impacts, new legal methodologies, legislative and administrative approaches, and innovative solutions must address ADCs’ water intake.
The project leads are Professor Marc Conte, an environmental economist in the Department of Economics, and Professor Kimberly Diamond, a transactional finance and energy lawyer in the Law School. Interested law students and undergraduate research assistants will help them develop potential solutions to the Issue by exploring two critical areas: (1) ADCs’ legal treatment, pre- and post-construction and (2) the economic analysis identifying how tradeoffs facing AI firms and society regarding the choice between open and closed-loop cooling systems vary depending on ACDs’ locations.