Preparing for AI Workloads: Options for Integrated Rack Power and Cooling Systems - BRKCOM-2530#
Cisco Live Europe 2026 (Amsterdam) digest Program Wildcard pick Amsterdam · 9–13 February 2026
Christopher Liljenstolpe (Cisco), Fabio Gori (Cisco)
- Why: Per-rack power planning jumping from 16-30 kW to 150-250+ kW in two years is the physical constraint that will reshape DC topology, cabling, and even optics choices — this is adjacent to the PM's physical-DC world but its consequences land squarely in it.
- What: The impact of rack power and thermal density choices on AI-ready facilities, how AI can optimize cooling energy and workload distribution across servers, and how to select cooling systems that prioritize energy efficiency without compromising performance (from abstract only).
- How: Options analysis across integrated rack power/cooling approaches, with AI-assisted optimization of cooling and workload placement.
- Where applicable: New AI build-outs and retrofits; the 150-250+ kW/rack figures are hyperscale/AI-cluster-specific. Liljenstolpe's DC background suggests this will be more grounded than a facilities-vendor pitch.