Google serves 3.2 quadrillion AI tokens a month — ~12GW, and energy could run out in 3 years
victor_explore · x · 2026-09-22
- Google's AI now serves 3.2 quadrillion tokens per month.
- AI chip founder Naveen Rao estimates that translates to roughly 12 gigawatts running nonstop — about a third of all US data center power.
- His projection: at this pace, energy supply becomes the binding constraint in about 3 years. Commenter victorexplore adds that power will stop being a sustainability slide and start capping how many agent calls per user you can afford.
More from Infra
- Same GPU, Four Ways to Buy It: What Nebius Spot Pricing Reveals About Compute Economics — demian_ai · 2026-09-22
- Tata Elxsi's IRIS: edge filtering cuts cloud frames 70-80% for real-time factory safety AI — AWS ML Blog · 2026-09-22
- bitsandbytes2 slightly delayed: zero-config lazy compression plus dynamic expert swaps for near-infinite KV cache — Tim_Dettmers · 2026-09-22
- Iterative sensitivity probing: how bitsandbytes2 finds each layer's compression limit — Tim_Dettmers · 2026-09-22
- Tim Dettmers releases runtime dynamic compression framework, hits 1.5-2.0 bit at high quality — Tim_Dettmers · 2026-09-22
- PSA: Non-US Users Should Consider Local AI in Case Governments Ban LLMs — TheMoonMidas · 2026-09-22