

177: AI And The Energy Grid: Solving for AI’s Power Needs
30 snips May 24, 2024
Will Su, a lead voice on energy at BlackRock, joins the discussion on the impending energy revolution prompted by AI advancements. He highlights the exponential rise in electricity demand required for AI technologies by 2030 and the crucial role of renewable energy sources like wind and solar in meeting these needs. Investment opportunities emerge in this shifting landscape, as traditional energy meets innovative solutions. Su also underscores the importance of evolving infrastructure to cope with the power surge, positioning the energy sector for transformation.
AI Snips
Chapters
Transcript
Episode notes
AI's Energy Consumption
- AI's high energy consumption stems from massive information processing.
- Computation involves moving electrons, and trillions of calculations add up quickly.
Power of LLMs
- Large language models (LLMs) are trained on trillions of parameters, requiring immense computational power.
- ChatGPT4, for instance, used 70,000 ZetaFLOPs, highlighting the scale of operations.
Projected AI Energy Demand
- AI's electricity demand could reach 1,000 terawatt hours by 2030, about 3% of global electricity.
- Total internet demand, including AI, might consume 6-7% of global electricity by 2030.