Epoch AI: LLM Query Energy Consumption Drops Below Human Brain's

Analysis by Epoch AI suggests that energy consumption per token for LLMs may have fallen below human brain levels. A single chatbot query now uses less energy than a microwave running for 10 seconds, though global AI compute demand continues to surge.

2026-07-21 ~ 2026-07-21 · 2 related posts

• LLM inference may already beat human inference on energy per token — jd_pressman · 2026-07-21
• Epoch AI says one chatbot query uses less energy than a 10-second microwave run — dl_weekly · 2026-07-21