

Solving the AI energy crisis (Changelog Interviews #652)
11 snips Jul 31, 2025
Greg Osuri, founder and CEO of Akash Network, shares his insights on the energy demands of AI and his recent congressional testimony. He discusses the urgent need for sustainable energy solutions to power AI technologies, including the potential of decentralized computing and renewable sources like solar and wind. Osuri highlights innovative strategies for harnessing excess energy and the importance of reforming outdated infrastructure. He also humorously touches on land affordability in Texas and the light-hearted scenario of a zombie apocalypse.
AI Snips
Chapters
Books
Transcript
Episode notes
AI's Looming Energy Crisis
- AI's energy demand is approaching a crisis, potentially consuming up to 12% of U.S. energy by 2028.
- The U.S. energy grid isn't expanding fast enough to meet AI growth, risking power shortages in data centers.
AI Data Centers and Environment
- Massive AI data centers currently burn fossil fuels due to grid limitations, worsening environmental health.
- Example: OpenAI's 4.5 GW capacity equates to the emissions of two thirds of Vermont's CO2 output.
Advocate Distributed AI Training
- Develop distributed training algorithms to spread AI compute workloads across multiple data centers and GPUs.
- This lowers energy concentration and enables elastic AI workloads that move to where energy is cheap and available.