AI queries got 33x more efficient, yet Google's power use rose 37%
FinanceYF5 · x · 2026-10-17
A 4-minute explainer on the efficiency paradox: Google says one AI query costs just 0.24 Wh, a 33x drop in a year, yet its total power consumption grew 37% last year. IEA projects global datacenter power use to double by 2030; GE Vernova gas turbine deliveries are booked into 2031; Texas grid's interconnection queue hits 474 GW, over 5x its historical peak. Efficiency isn't a brake—it's an accelerator (Jevons paradox).
More from Infra
- NeodraDynamics unveils Neo1, a 4-meter 1.2-ton heavy-duty humanoid rated for 300 kg per arm — CyberRobooo · 2026-10-17
- Replit Launches BYOC: Run Dev and Production in Your Own Cloud on AWS, Azure or GCP — amasad · 2026-10-17
- From 15 to 60+ tokens/s: a year and a half of llama.cpp on dual RTX 5060 Ti — Kahvana · 2026-10-17
- ModCon 2026 panel: silicon locks in years ahead while software ships weekly — clattner_llvm · 2026-10-17
- Darkbloom Turns Your Mac Into a Paid AI Inference Server — gajesh · 2026-10-17
- ExecuTorch is now an official Zephyr module, bringing PyTorch models to MCUs via west — PyTorch · 2026-10-17