Live Panel: Building at the Roofline — Planning and Delivering Gigawatt-Scale AI Infrastructure
summary
Plans target gigawatt-scale fleets with company capex comparable to a national GDP and a planning model split into two planning cycles (0–18 months and 3–5 years) with long lead reservations, co‑investments and supplier roadmaps. Surprises include an Earth-to-moon worth of fiber for a 300k‑GPU cluster, retrofitting a 24k→100k GPU site from InfiniBand to Rocky, amplifier‑hut space/power re‑designs, and constrained t-minus-3 BOM resiliency for wafers, HBM and advanced packaging that forces multi‑vendor interoperability and pre‑integrated racks.
AI-written summary · about
chapters (8)
- 0:00Gigawatt AI infrastructure
- 1:29From just-in-time to long-term
- 5:23Scale challenges in AI builds
- 9:19Supply challenges and solutions
- 15:02Multi-vendor interoperability
- 21:05Co-development partnerships
- 23:13Pre-integrated racks challenges
- 26:01Closing remarks
related talks
24mClosing Keynote
18mThe Geopolitics of AI Infrastructure
14mFireside
28mScaling agentic AI: From infrastructure to enterprise outcomes
52mKeynote Address: Insights Gained from Delivering Two Generations of AI Supercomputers...
01h 13mEngineering the Cloud for Tomorrow: Performance, Speed and Scale for AI Workloads
09mDemocratizing Access to GPUs for Faster MLAI Innovation with Generative AI- 28mScaling agentic AI: From infrastructure to enterprise outcomes – Kevin Cochrane