The Next Evolution of Kubernetes: GPU-Centric Infrastructure for AI Workloads
summary
Fujitsu presents a shift from CPU-first clouds to GPU-centric infrastructure that dynamically composes hardware to match AI demand, claiming Codi (open source) can scale hardware up and down to cut GPU, memory and electricity waste. The stack pairs software-led orchestration with Composable Disaggregated Infrastructure, and Codi is now a CNCF Sandbox project; the demo-level claim is resource and power optimization for unpredictable Agentic AI workloads rather than a completed production rollout.
AI-written summary · about
related talks
05mTowards a Sustainable World Realized by Kubernetes and Composable...
15mKeynote: From Platforms to AI Factories. Has Kubernetes Solved It?
30mRunning AI Workloads in Containers and Kubernetes
05mSponsored Keynote: Towards Clouds of AI Clusters
36mNow we’re all cloud natives, what’s next?
40mPanel: Kubernetes and GPU Trends for AI in Financial Services
20mOperating Frontier Intelligence at Scale
38mBeyond Kubernetes: Adapting To Spe...