Cost of AI: Impact of Artificial Intelligence on the Environment

Abstract

This paper explores the environmental cost of artificial intelligence (AI). Exploring the significant resource used in AI dataceners. This papers examines the environmental cost of AI through a detailed case study of xAI’s “Colossus” supercomputer cluster in Memphis, Tennessee—the world’s largest single-site AI training facility as of 2025, reportedly housing over 100,000 NVIDIA H100/H200 GPUs and projected to scale toward 1 million GPUs. Using publicly available utility records, satellite imagery, permitted emissions data, plans release by xAI and reported power and water agreements, this paper tries to quantiy  the facility’s footprint across multiple environmental dimensions.

Results indicate that Colossus currently consumes approximately 150–180 MW of continuous power, with approved expansion pathways reaching 500+ MW—making it one of the single largest industrial electricity loads in the southeastern United States. AI requires tramendous water useage for its cooling requirements, which drive extraordinary water usage, estimated at 1.2–1.8 million gallons per day under current operations, primarily sourced from the Memphis Aquifer, with significant wastewater discharge containing elevated temperatures and trace chemicals into local sewer systems. Air-quality impacts stem from backup diesel generators (and legal loopholes currently being exploited by xAI) and upstream emissions from TVA’s natural gas and coal-heavy grid, contributing an estimated 600,000–800,000 metric tons of CO₂-equivalent emissions annually at full Phase-I buildout. Local particulate Formaldhyde and NOx increases have been documented within a 10-km radius since commissioning in 2024.This case study illustrates the true cost of AI from an environmental perspective.

Prathyaj Mantha is a Systems Engineer Consultant at PEER Consultants P.C in Washington and a current MBA student at the University Of Virginia -Darden School Of Buissness. With over 5 years of experience in the systems engineering space, he has authored papers on a variety of environmental issues from a system engineering perspective.