Everything Energy How energy will shape the future of AI (and vice versa)
16 snips
May 15, 2025 Thomas Spencer and Siddharth Singh, lead authors of the IEA's report on Energy and AI, share their insights on the burgeoning relationship between artificial intelligence and energy consumption. They discuss how the demand for electricity from data centers will surge by 2030 and the dual nature of AI—while it can enhance energy efficiency, it also raises concerns about carbon emissions. They highlight the vital role of supply chains and renewable energy in shaping this dynamic, and how AI is transforming the very fabric of our energy systems.
AI Snips
Chapters
Transcript
Episode notes
ChatGPT Made AI Energy A Mainstream Issue
- ChatGPT's public launch made AI a mainstream commercial technology with big energy implications.
- The IEA report links AI's rapid rise to substantial electricity demand growth in data centres.
Large Models Drive Disproportionate Energy Use
- Training large language models has become energy-intensive; some trainings consumed ~2 TWh historically.
- A single reasoning model query can use 10–20x more electricity than small-model queries.
IEA Reconstructed Data Centre Energy Use
- No comprehensive global statistics exist for data centre electricity; the IEA combined multiple data sources to estimate consumption.
- They used counts of facilities, company reports, national stats, and industry interviews to build global estimates.