Kubeflow 正式从 CNCF 毕业，NVIDIA 与 AWS 等厂商共话云原生 AI
做 MLOps 的朋友可以关注下：Kubeflow 正式从 CNCF 毕业了，NVIDIA、Apple、AWS 的人都在台上给它背书。如果你的训练任务跑在 Kubernetes 上，配合 PyTorch 和 DeepSpeed 这套组合的稳定性又多了一层保证。
Kubeflow 在 KubeCon + CloudNativeCon 大会上宣布从 Cloud Native Computing Foundation（CNCF）毕业。来自 NVIDIA、Apple 与 AWS 的五位工程师 Ron Kahn、Andrey Velichkevich、Vara Bonthu、Ekin Karabulut 和 Manabu McCloskey 参与了这场关于云原生 AI 下一阶段的讨论。
演讲者特别提到 Kubeflow 的 CNCF 毕业里程碑，并点名 PyTorch Foundation 旗下的 PyTorch 与 DeepSpeed 两个项目。他们的核心观点是：跨基金会协作对维持健康的开源生态至关重要，Kubeflow 作为 Kubernetes 上的机器学习工作流平台，正是连接云原生社区与 AI 训练框架的枢纽。
Kubeflow 毕业意味着项目已达到 CNCF 对成熟度、治理和采用度的最高标准，与 Prometheus、Kubernetes 等处于同一层级。对企业用户来说，在生产环境中用 Kubeflow 编排 PyTorch、DeepSpeed 训练任务的技术栈稳定性得到进一步背书。
Ron Kahn of @nvidia, Andrey Velichkevich of @Apple, Vara Bonthu of @awscloud, Ekin Karabulut of @nvidia, & Manabu McCloskey of @awscloud discuss powering the next era of cloud native AI with an open community at #KubeCon + #CloudNativeCon today, citing @kubeflow's @CloudNativeFdn graduation with shout outs to PyTorch Foundation projects - @PyTorch and @DeepSpeedAI. The importance of working across foundations is vital for a thriving ecosystem.
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Kubeflow, CNCF, PyTorch, DeepSpeed.
Source attribution: curated from PyTorch (@PyTorch). Cross-source corroboration: 11 independent sources have covered this development (including CoreWeave, NVIDIA AI, Sebastian Raschka), confirming industry-wide signal strength.
