Karmada has officially graduated from CNCF and is now used for multi-cluster AI training and GPU scheduling

On September 8, 2026, the Karmada project officially graduated from the Cloud Native Computing Foundation (CNCF). As a multi-cluster management tool in the Kubernetes ecosystem, Karmada aims to solve the problem of resource coordination across clusters and supports unified orchestration of workloads and resources between multiple Kubernetes clusters. Its core features include a global view, policy-driven deployment, automatic scaling, and fault recovery, effectively improving the management efficiency of large-scale heterogeneous clusters. With Karmada’s graduation, CNCF has placed it on a separate project maintenance track, focusing in the future on enhancing multi-cluster management capabilities and ecosystem compatibility. Currently, Karmada has been practically applied in multi-cluster AI training and GPU scheduling scenarios.