The Hidden Energy, Water, and E‑Waste Footprint of AI

In 2023, data centres consumed 4.4% of all US electricity, a figure that could triple by 2028 as AI workloads surge. A 2021 Google‑UC Berkeley study estimated that training GPT‑3 alone consumed 1,287 megawatt hours of electricity and produced 552 tonnes of CO₂, enough to power roughly 120 average American homes for a year. Deploying models in real‑world applications and fine‑tuning them draws large amounts of energy long after the initial development phase. A single ChatGPT query uses nearly ten times the electricity of a standard Google search, and billions of queries per day amplify this impact. AI servers across the United States could generate an annual water footprint of 731 to 1,125 million cubic metres from 2024 to 2030, exceeding bottled water consumption worldwide in a single year. The same expansion is expected to produce 32.6 to 79.7 million tonnes of CO₂ each year, with only 22% of global e‑waste properly collected and recycled. AI‑specific hardware becomes obsolete within two to three years, contributing 1.2 to 5 million metric tonnes of e‑waste by 2030, while the world already generated 62 million tonnes of e‑waste in 2022. States like Virginia and South Carolina will see AI data centres consume 36–51% and 65–70% of their total electricity by 2030, respectively, straining local power and water supplies.