The Next Evolution of Kubernetes: GPU-Centric Infrastructure for AI Workloads - Takao Indoh, Fujitsu

This talk examines how Kubernetes is adapting to serve as a GPU-centric orchestration platform for generative AI and Agentic AI workloads. It addresses critical challenges including explosive computational demand, global GPU shortages, and rising energy costs that strain traditional cloud infrastructure. The presenter demonstrates Fujitsu's Composable Disaggregated Infrastructure approach as a solution for dynamically allocating and optimizing expensive GPU resources across AI workloads.