Cloud Native Computing Foundation Announces Karmada Graduation

Karmada's graduation marks its transition from Incubating status, which it held since December 2023, to the CNCF's top maturity level. The project entered CNCF as a Sandbox project in September 2021, with its first commit dating to November 2020. To reach graduation, the project completed a third-party security audit, established a formal steering committee, adopted the CNCF Code of Conduct, and maintains a CII Best Practices Badge. These are the governance and security milestones the CNCF Technical Oversight Committee requires before endorsing a project as enterprise-ready.

The technical proposition is multi-cluster Kubernetes orchestration without application modifications. Karmada extends the standard Kubernetes API with centralized placement, propagation, failover, and multi-cluster autoscaling, meaning teams can coordinate workloads across clusters, clouds, and regions from a single control plane. It integrates with existing CNCF tooling: it exports Prometheus metrics from control-plane components, packages an etcd instance for control-plane state, and ships Helm charts for installation. For teams already running multiple Kubernetes clusters, this addresses the operational gap that native Kubernetes does not solve on its own.

The v1.19 release and the broader 2026 roadmap point toward AI workload scheduling as the primary growth area. Version 1.19 advances multi-component scheduling for distributed AI training jobs and promotes priority-based scheduling to Beta, enabled by default, so critical workloads are scheduled first. The roadmap adds priority-based preemption, multi-cluster queuing for AI training and batch jobs, and multi-cluster support for Kubernetes Dynamic Resource Allocation across GPUs and other accelerators. This positions Karmada as a coordination layer for GPU-constrained environments where capacity is spread across clusters and providers.

The adopter base skews heavily toward Chinese cloud, internet, telecom, and AI companies, including Alibaba Cloud, Bilibili, Huawei, iFLYTEK, JDCloud, Kuaishou, RedNote, SenseTime, Trip.com, Vivo, WPS, and ZTO. Global adopters named are Bloomberg and Wellhub. The project reports more than 1,214 contributors across 292 contributing organizations and more than 5,600 GitHub stars. The concentration of adoption in China is worth noting for teams evaluating the project: the production evidence is strong but geographically uneven, and the contributor base reflects that distribution.

Only one feed carried this story, which limits corroboration. The announcement was made at KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026 in Shanghai. Because the source is the CNCF's own announcement, the framing is naturally promotional, and independent assessment of the project's production readiness at scale outside the named adopters is not available from this material. Teams considering adoption should weigh the graduation criteria, which are verifiable, against the geographic concentration of reported production usage.