Sustainable & Smart AI Factories in Europe
summary
Panelists call the €100B CERN for AI idea ambitious but under-specified, noting concrete bottlenecks: supply lag for B200 GPUs, spine‑leaf networking changes, cooling and energy timing (the Netherlands hit 150% renewables daytime last summer), and fragile inter‑GPU fabric that breaks training at scale. They point to real builds — campus‑scale clusters, Leonardo training runs, a Barcelona supercomputer conversion — and push federation, predictive AI for maintenance, and data‑center grid services (batteries/hydrogen) as practical fixes.
AI-written summary · about
chapters (11)
- 0:00AI data center panel intro
- 5:25Europe’s AI investment critique
- 10:45European AI infrastructure challenges
- 17:00Government vs private sector gaps
- 22:59AI sovereignty policy & infrastructure
- 26:52AI infrastructure ROI & adoption
- 31:51Sustainable AI innovation balance
- 37:28Network reliability & data transfer
- 40:43Q&A energy flexibility & funding
- 43:58Germany AI funding perspective
- 48:55Europe’s AI future closing
related talks
02mStarcluster
46mChallenges of Making Large AI Clusters Reliable
17mFocus on Data, on the Road to AI
05mAI and ML are changing traditional business
35mBuilding a 5-Exaflop Supercomputer for Meta-AI Research and...
16m[Opening Keynote] Making Distributed Computing Easy
03mData Fueling AI of the Future
24mHow to Drive Impact and Scale with Your Big Data/AI Applications?