NVIDIA adds runtime AI inventory to its open source Kubernetes cluster tool
NVIDIA's AI Cluster Runtime (AICR), an open source component of its DSX AI Factory platform, now includes a qualified AI bill of materials component built on k8s-aibom, the project Google created to watch AI workloads on Kubernetes. Two stock GKE H100 recipes select it by default, and other GKE recipes can turn it on with a --runtime-inventory flag. NVIDIA reports roughly four API-server requests per minute and 2.2 millicores average CPU while observing. Inventory is opt-in per namespace, and the code is in both repositories.
Why it matters
Teams can now pair a version-locked GPU platform configuration with a record of which models and serving runtimes are actually running on it, which the source ties to security reviews, incident response and drift detection. Reporting stays opt-in per namespace, so nothing leaves the cluster by default.
Sources
- PrimaryNVIDIA Developer
Written by AIToolMaps from the source above and checked against it sentence by sentence.