The impact of AI on the environment

Abstract

This paper provides a quantitative and technical analysis of the physical infrastructure supporting large-scale AI models like GPT-4. Rather than focusing on general productivity benefits, the study examines the critical surge in power density within data centers and predicts their emergence among the world's top five electricity consumers by 2026. Special attention is paid to the «water footprint» of computation in resource-stressed regions and the ecological degradation caused by the extraction of rare-earth minerals for GPU production. The author argues for a transition from corporate secrecy to mandatory transparency metrics and algorithmic optimization as fundamental requirements for environmental safety in the AI era.