KubeCon China Schedule Drops: AI Agents, GPU Virtualization, and Meta's 100k-GPU Fault Tolerance

The schedule for KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China is now available, setting the stage for September 7-9 in Shanghai. The combined event brings together engineers, maintainers, and researchers from the cloud native, open infrastructure, and PyTorch ecosystems. Key sessions include Meta's approach to keeping 100k-GPU jobs alive with open-source fault tolerance (presented by Tristan Rice and Amir Afzali), GPU virtualization at scale using HAMi (from dynamia.ai and Intsig), and a look at how AI agents are rewriting the OpenStack/K8s playbook (DOCOMO Euro-Labs and NTT DOCOMO). Other notable talks cover Kata Containers 4.0 as a sandbox for the agent era, the state of the PyTorch ecosystem in 2026, and a controlled AI agents approach for scaling digital employees at China Merchants Bank.

In addition to the main conference, two co-located events will run on September 6-7. OSPOlogy + OSPO Summit China (Sept 7) focuses on the evolution of Open Source Program Management and corporate governance in the age of Agentic AI. AGNTCon + MCPCon China (Sept 6-7) concentrates on building reliable, scalable, and secure agent systems in practice. Registration for all events is open on the official website. This conference offers a concentrated view of how open-source infrastructure is enabling AI production workloads, particularly for professionals working in cloud native, AI ops, and large-scale GPU clusters.