Kubeflow graduates within the CNCF, nine years after its creation at Google

2026-08-17 · Publisher : CNCF · Domain : Cloud & DevOps

On 17 August 2026, the CNCF announces the graduation of Kubeflow, the open source AI and ML lifecycle platform on Kubernetes: nearly 260 million PyPI downloads, more than 6,600 contributors from over 1,000 organizations.

### Background On 17 August 2026, the CNCF announces the graduation of **Kubeflow**, an ecosystem dedicated to standardizing data and AI workloads on Kubernetes, from data processing to distributed training, fine-tuning and inference. Created at Google in 2017, Kubeflow joined the CNCF in incubation in 2023; the CNCF describes an evolution from a collection of components to a unified AI-native platform over about nine years. ### Key facts - Nearly 260 million PyPI downloads of Kubeflow Python packages. - More than 33,000 GitHub stars, more than 6,600 contributors and more than 1,000 organizations. - Scope: data processing, interactive development, distributed training, fine-tuning, inference and model serving, across public, private and hybrid cloud. - Users cited: Bloomberg, NVIDIA, Red Hat, LinkedIn and Spotify. - Co-founder quoted: David Aronchick; the CNCF presents graduation as a maturity signal for enterprise AI workloads. ### Why it matters Kubeflow is one of the oldest AI building blocks of the Kubernetes ecosystem, and its graduation follows KServe's incubation and llm-d's Sandbox entry: the CNCF is assembling a reference stack for AI, from training to inference. For an IT services firm, it is a longevity argument against proprietary MLOps platforms. The platform remains complex to operate: Graduated attests to the project's maturity, not to how easy it is to run. ### Takeaway The CNCF now has a reference open source AI stack on Kubernetes, with Kubeflow at the top maturity level.

Tags : Kubernetes, IA, Open Source, Cloud natif, Données