Prefer to listen? Here's the narrated myth-vs-reality briefing.
Listen to Energy and environment
Plus unlocks audio streaming. Pro adds downloadable audio, video, certificates, and more.
The Myth
The "AI is destroying the climate" framing often quotes worst-case projections and ignores rapid efficiency gains. Inference costs per token have dropped roughly 10-fold each year for several years. Major hyperscalers (Microsoft, Google, Amazon) have committed billions to nuclear, geothermal, and renewable power purchase agreements specifically for AI workloads.
Sources:IEA Electricity 2024 (data centers)Epoch AI compute trends
The Reality
Data center electricity demand IS growing fast — a July 2026 BloombergNEF forecast projects US data centers will draw nearly 200 gigawatts by 2035, about 20 percent of the nation's electricity (up from under 6 percent today and roughly four times current demand), an estimate 83 percent higher than the same firm's December projection. Cooling water draws are non-trivial in water-stressed regions. The grid has not kept pace with hyperscaler buildout in some markets. These are legitimate environmental considerations, not paranoia.
The Positive Path
AI is also one of the most powerful tools we have for the *energy transition itself*: optimizing grid operations, designing better batteries and solar cells, accelerating materials science, and forecasting renewable generation. DeepMind's grid-cooling work cut Google's data center cooling energy by 40%. The bet that "AI helps decarbonize faster than it consumes" is plausible — but only if we measure and act on both sides.