KQED's Forum California’s AI Data Centers Taking Growing Environmental Toll
Dec 18, 2025
Molly Taft, a senior climate reporter at WIRED, and Aaron Cantú, a staff writer at Capital & Main, dive into the environmental impact of California’s data centers. They discuss the alarming energy demands these facilities have due to AI, leading to increased fossil fuel emissions. Taft highlights the complex relationship between data centers and water use, while Cantú emphasizes the strain on local grids. They also explore community pushback and the lack of transparency in energy consumption, revealing a deeper concern about the future of energy policy amid the AI boom.
AI Snips
Chapters
Transcript
Episode notes
AI Workloads Amplify Resource Needs
- Data centers shifting to AI workloads are much larger and demand far more electricity and sometimes water than older centers.
- Hyperscaler campuses from Meta, Google and others change resource needs over the next 5–10 years.
Water Cooling Trades Electricity For Water
- Many data centers use potable water for evaporative cooling because salt and impurities would corrode equipment.
- Choosing water cooling lowers electricity but raises local water stress and trade-offs with chemicals and forever-chemical coolants exist.
One-Number Estimates Are Misleading
- Published water-use figures (like OpenAI's) lack standardization and omit key context like training vs inference or offsite electrical water impacts.
- Water and energy footprints are site-specific and often proprietary, so single global figures mislead.