Finshots Daily The AI crisis no one’s talking about
18 snips
Feb 10, 2026 They dig into how hyperscale AI campuses demand city-scale power and why per-request energy for chat models dwarfs a simple search. They cover slow grid upgrades versus rapid AI buildouts and the IEA’s staggering electricity projections. They describe jet engines repurposed as quick power, the health and climate trade-offs, and tech firms turning into power producers while eyeing small modular reactors.
AI Snips
Chapters
Transcript
Episode notes
AI Centers Are Massive Power Consumers
- AI data centers demand concentrated, continuous power comparable to mid-sized cities.
- This surge can outpace grid upgrades because AI infrastructure scales in months while grids take years.
Chat Requests Scale To Huge Electricity Loads
- A single chat request uses far more energy than a typical search and scales to enormous annual loads.
- IEA estimates show AI electricity demand jumping from 8 TWh in 2024 to 652 TWh by 2030, an ~80x rise.
Jet Engines Reborn As Data Center Generators
- Companies are repurposing retired jet engines as fast-deployable gas turbines to power data centers.
- Firms like Crusoe Energy have ordered aviation turbines to provide off-grid power near AI facilities.