The Real Cost of a Token: the Hidden Bill of Generative AI
summary
The talk quantifies the environmental cost of generative AI beyond token or latency metrics, measuring energy (kWh), water use and e‑waste from model training and inference and comparing LLM requests to traditional search. It analyzes concrete data on training and inference footprints and then presents practical solutions—right‑sizing architecture, optimizing pipelines, and adopting carbon‑aware practices—to build more efficient, economically and environmentally sustainable AI.
AI-written summary · about
chapters (12)
- 0:00Hidden AI bill intro
- 1:04Generative AI growth
- 2:30Training energy costs
- 4:36Search engine energy
- 6:00Jevons paradox
- 7:03Data center water use
- 8:26GPU/CPU resource use
- 9:08Eco-friendly LLMs
- 10:12Prompt optimization
- 11:14Local hosting and SLMs
- 12:18Eco-responsible autoscaling
- 13:03Sustainable AI impacts
related talks
01h 02mJoint Keynote Address: Scaling AI Sustainably: An Uncharted Territory- 01h 02mJoint Keynote Address: Scaling AI Sustainably: An Uncharted Territory
06mThe Dark Side of AI - Environmental & Social Costs Developers Must Know
41mSustainable Digital
17mGenerative AI: Keeping it real
47mThe Rise of AI: A Reality Check on Generative AI
20mBuilding GenAI That Doesn't Cost the Earth
38mArtificial Intelligence: How Much Will It Cost You?