Particle.news

Download on the App Store

Generative AI Drives Surge in Data Center Energy Use and Emissions

A new report quantifies AI's energy demands, revealing per-query costs and forecasting a doubling of global data center electricity consumption by 2030.

Image
Image
Image
Image

Overview

  • The MIT Technology Review report details per-query energy use, ranging from 114 joules for smaller models to 6,706 joules for larger ones like Llama 3.1 405B.
  • AI-generated video is highly energy-intensive, with a five-second clip consuming about 3.4 million joules, far exceeding the energy demands of text or image generation.
  • A heavy AI user could consume approximately 2.9 kWh daily, equivalent to running a microwave for over three hours.
  • U.S. data center electricity use has doubled since 2017, with projections showing half of this power will support AI tools by 2028.
  • Global data center energy consumption is expected to more than double by 2030, surpassing Japan's total energy use and significantly increasing greenhouse gas emissions.