A Harvard-led study estimates that 403 hyperscale data centers in the US consumed approximately 82 terawatt-hours of electricity in a year, representing about 1.8% of the nation's total power usage. This consumption generated an estimated 25.7 million metric tons of CO2, a figure two to three times higher than a 2018 estimate. The study noted that over half of this energy was derived from fossil fuels, primarily gas and coal, though the carbon intensity per kilowatt-hour was slightly lower than the national average. AI
IMPACT Highlights the substantial energy and environmental footprint of AI infrastructure, potentially influencing policy and development.
RANK_REASON The cluster reports on a study detailing the significant energy consumption and CO2 emissions of hyperscale data centers, a key infrastructure component for AI.
Read on Mastodon — mastodon.social →
- Mastodon
- United States
- Baxtel
- Environmental Protection Agency
- Harvard University
- Michigan
- Oregon
- PJM Interconnection
- US
- Wired
- Ars Technica
AI-generated summary · Google Gemini · from 4 sources. How we write summaries →