CoreWeave Moves to CNCF’s Platinum Tier as AI Inference Work Expands

CoreWeave upgraded to the Cloud Native Computing Foundation’s top membership tier, a move that coincides with its engineers’ work to prepare open-source inference software for newly released AI models.

CoreWeave announced its move to Platinum on Tuesday, Nov. 10, during KubeCon + CloudNativeCon North America in Salt Lake City. The event ran Nov. 9–12.

CoreWeave engineers contribute to Kubernetes, Kueue and CloudNativePG. The company also participates in CNCF’s Certified Kubernetes AI Conformance Program and was a founding contributor to llm-d alongside Red Hat, IBM, Google and NVIDIA. llm-d is a CNCF Sandbox project.

Kubernetes is open-source software for managing containerized applications, and the CNCF hosts the project. CoreWeave has built its cloud on Kubernetes and uses the platform to run AI models at production scale.

Platinum is the CNCF’s highest membership tier. Its published terms include a $350,000 fee, a Linux Foundation membership and a minimum three-year commitment. CoreWeave’s specific payment was not disclosed.

Separately, llm-d’s version 1.0 release on Oct. 20 added workload controls and routing features for running large models on Kubernetes. The project distributes inference requests among model servers such as vLLM and SGLang. Inference is when a deployed model processes input to produce an output.

Rita Zhang, a principal software engineer at CoreWeave, said engineers work with model labs and open-source projects including vLLM and SGLang to patch inference engines and runtimes. They also provide company-built kernels for “day-zero” models, or models that have just been released.

“Just because it’s on Hugging Face and you download the weights doesn’t mean it will work per your requirement,” Zhang said.

Shadi Saba, CoreWeave’s vice president, engineering, AI/ML, said the company is applying lessons from its operations to “large-scale training, production inference, and agentic workloads.”