NVIDIA AI Updates: November 12, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. NVIDIA AI Cluster Runtime Adds Runtime AI Inventory, Emitting CycloneDX AI Bills of Materials for Running Kubernetes Workloads
NVIDIA described Runtime AI Inventory on November 11, a component of AI Cluster Runtime (AICR), the open source module of the DSX AI Factory Platform that produces version-locked recipes for GPU Kubernetes clusters. It builds on k8s-aibom, a Google-created project that watches Kubernetes workload and pod APIs and emits CycloneDX 1.6 AI bill of materials records for what is actually running. Each attribute carries one of four confidence tiers (declared, inferred, unresolved, verified) and an evidence locator. Covered workload kinds include native Dynamo graphs, NIMService, LeaderWorkerSet, DisaggregatedSet and CronJob, and inventory is opt-in per namespace. Two stock recipes (h100-gke-cos-inference and h100-gke-cos-inference-dynamo) enable it by default, and other GKE recipes opt in with --runtime-inventory; the validated environment is GKE 1.32 or later on H100 nodes. In upstream k8s-aibom v1.6.1 testing the controller averaged 2.2 millicores of CPU and 28 to 34 MiB of memory. NVIDIA notes that runtime observations are unsigned, Sigstore verification is off by default, and correlating recipe intent with observed workloads is future work. Licenses for the two repositories are not stated in the post. Source