# NVIDIA Links GPU Deployment Recipes With AI Workload Inventory By Simon Yoon Canonical URL: https://www.tokenpost.com/news/technology/55105 Published: 2026-11-12T07:55:19.000Z Updated: 2026-11-12T07:55:19.000Z Section: Technology > AICR packages validated GPU configurations, while Google-developed k8s-aibom inventories recognized models, runtimes and dependencies in Kubernetes workloads. NVIDIA announced an integration that pairs GPU deployment guidance with an inventory of AI workloads running on Kubernetes. The company’s AI Cluster Runtime (AICR) prepares validated platform configurations, while Google-developed k8s-aibom records details about recognized workloads. Together, the tools address platform setup and runtime visibility within NVIDIA’s DSX AI Factory Platform. AICR generates version-locked recipes and reproducible deployment bundles. Each recipe defines the hardware, cloud, operating system and workload intent. AICR checks configurations against deployment paths and combines upstream sources, charts and controller images pinned by digest into tested sets, with security defaults reviewed. Bundles include integrity checksums, can be deployed with Helm, Flux or Argo CD, and may be signed with the aicr bundle --attest command. The runtime inventory produces CycloneDX 1.6 AI bills of materials, or AIBOMs, listing models, inference runtimes and dependencies. k8s-aibom recognizes resources including Dynamo graphs, NIMService, LeaderWorkerSet, DisaggregatedSet and CronJob. The h100-gke-cos-inference and h100-gke-cos-inference-dynamo recipes enable it by default. Other Google Kubernetes Engine (GKE) recipes can activate it with the --runtime-inventory flag. During steady-state operation, k8s-aibom makes about four requests per minute to the Kubernetes API server. Its average CPU use is 2.2 millicores, and its memory working set ranges from 28 to 34 MiB. The inventory is intended to help teams review security, meet compliance requirements and respond to incidents. It can also help surface unregistered workloads, dependency changes, external AI APIs and vector databases. The integration follows NVIDIA’s Nov. 10 public preview of [two Kubernetes tools for GPU clusters](). NVIDIA Kubernetes Engine (NVKE) handles cluster provisioning and management; NVIDIA Kubernetes Recipes (NVKR) builds software images. The preview placed both tools within the DSX platform. ## Links in this article - [two Kubernetes tools for GPU clusters](https://www.tokenpost.com/news/technology/54177)