{"resourceId":"nvidia-aicr-vera-rubin-nvl72-kubernetes-preview-recipes-20261102","versions":[{"version":"external-e1c4c959232b1a856a746defcf560b286b9241c159f8688e8ba5e3e08793e2e3","resource":{"id":"nvidia-aicr-vera-rubin-nvl72-kubernetes-preview-recipes-20261102","title":"NVIDIA's AICR adds Preview Kubernetes recipes for Vera Rubin NVL72, validated on early hardware, and lists the driver, NIC and topology changes from GB200","organization":"NVIDIA","sector":"AI infrastructure / Kubernetes platform software","geography":"Global","publishedAt":"2026-11-02","publicationDate":"2026-11-02","eventDate":"2026-11-02","sourceName":"NVIDIA Technical Blog","sourceLabel":"NVIDIA blog: Making Kubernetes AI-Ready for NVIDIA Vera Rubin on Day One","sourceUrl":"https://developer.nvidia.com/blog/making-kubernetes-ai-ready-for-nvidia-vera-rubin-on-day-one/","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["infrastructure","developers-agents"],"finding":"NVIDIA's AI Cluster Runtime (AICR) adds a Vera Rubin NVL72 recipe family (RKE2, Ubuntu, training, inference and Dynamo variants), validated end to end on early Vera Rubin clusters with GPU Operator v26.7.0, DRA driver v0.5.0 and a host-installed R615.23 driver. The post lists changes from GB200 NVL72: a new driver branch (R615.23 tested, not a confirmed minimum), ConnectX-9 SuperNICs and BlueField-4 DPUs replacing ConnectX-7 and BlueField-3, a different GPU-to-NIC topology, new MIG profiles not covered by the validation, and Kubernetes 1.34.1 or later. Early validation was blocked by a hardware/fabric issue that kept one tray out of the NVLink fabric until reseated. The recipes ship as Preview, without the production support and lifecycle qualification of Supported recipes.","sledRelevance":"Relevant to state, university and national-lab-adjacent buyers planning next-generation NVIDIA systems: it shows Vera Rubin will require new drivers, NICs, DPUs and Kubernetes versions, and that the software path is still pre-release. Useful for procurement timing, not for current deployments.","evidence":"Vendor validation on early-access reference clusters; one bus-bandwidth gate reported. Preview status stated by NVIDIA.","architectureImplications":"Upgrading from GB200 means new driver branch, ConnectX-9/BlueField-4 networking, NUMA/NIC affinity checks and DRA-managed IMEX.","governanceImplications":"Preview recipes should not be accepted as production baselines in contracts.","securityPrivacyImplications":"BlueField-4 DPUs replace BlueField-3; DPU security posture needs fresh review (BlueField DPU tampering research was covered in an earlier edition).","caveats":"Pre-release platform and Preview recipes; MIG not validated; tested versions are not stated minimums.","streamIds":["nvidia"],"roles":{"sales":"Use this to set realistic expectations with public buyers considering Vera Rubin. NVIDIA itself labels the Kubernetes recipes Preview, and lists many changes from GB200: drivers, NICs, DPUs, topology and Kubernetes versions. For a university or state facility planning a 2027 purchase, offer a readiness review of their current GB200 or Hopper stack against these changes. Do not suggest Vera Rubin is production-ready on Kubernetes today, and do not claim performance beyond the single bandwidth gate NVIDIA reports from early hardware.","engineering":"Plan the delta from GB200 now. Expect a new driver branch (R615.23 tested, R620 for official support), ConnectX-9 and BlueField-4 replacing ConnectX-7 and BlueField-3, changed GPU-to-NIC affinity, DRA-managed IMEX with the host imex service disabled, and Kubernetes 1.34.1 or later. MIG on Rubin is not validated. Proof of value: run aicr validate on the current cluster, document gaps against the Vera Rubin recipe, and re-check when Supported recipes ship. Review BlueField-4 DPU security separately.","delivery":"Treat Vera Rubin adoption as a staged platform upgrade. The platform engineering team owns recipe versions and validation results. Schedule a readiness assessment, track NVIDIA's move from Preview to Supported recipes, and set contract acceptance on Supported status. Train staff on DRA, ComputeDomains and IMEX. Risks include early-silicon hardware issues like the fabric fault NVIDIA hit, version churn before GA, and untested MIG behaviour for multi-tenant sharing."},"roleDetails":{"sales":{"customerProblem":"Public buyers planning next-generation GPU purchases need to know how ready the software stack is.","stakeholders":["Research computing director","State IT infrastructure lead","Procurement officer"],"discoveryQuestions":["Are you planning Vera Rubin in your next refresh?","Do you run Kubernetes or Slurm on current GPUs?","Will you need MIG for multi-tenant sharing?"],"valueHypothesis":"An early readiness review reduces surprises when Vera Rubin ships.","potentialEngagement":"Readiness review of a current GPU Kubernetes stack against AICR Vera Rubin recipes.","unsupportedClaims":["Production readiness","Performance beyond the single reported gate","MIG support"]},"engineering":{"fit":"Relevant for Kubernetes-based GPU platforms planning Vera Rubin NVL72.","architectureIntegration":"AICR recipes generate bundles for GPU Operator, DRA driver with ComputeDomains, container toolkit and DRANET.","prerequisites":["Kubernetes 1.34.1+","Ubuntu with 64k-page kernel","R615-branch driver","ConnectX-9 / BlueField-4 networking"],"constraints":["Preview status","MIG not validated","Tested versions not confirmed minimums"],"security":"New BlueField-4 DPUs need their own security review.","proofOfValue":["Run aicr validate on the current cluster and log gaps","Re-validate when Supported recipes ship"]},"delivery":{"work":"Readiness assessment and staged upgrade plan for Vera Rubin.","dependencies":["NVIDIA Supported recipe release","Hardware delivery","Driver GA"],"ownership":"Platform engineering team.","skillsAdoption":"Training on DRA, ComputeDomains and IMEX.","governanceCheckpoints":["Acceptance tied to Supported recipes","DPU security review"],"acceptanceCriteria":["Proposed: aicr validate passes on delivered hardware with Supported recipe","Proposed: NCCL bandwidth gate met at acceptance"],"risks":["Early-silicon hardware faults","Version churn","Untested MIG"]}},"retrievedAt":"2026-11-03T04:00:36Z","enrichedAt":"2026-11-03T04:04:41Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Platform teams need Kubernetes DRA and IMEX skills.","procurementImplications":"Buyers should tie acceptance to Supported, not Preview, software status and confirm driver branches at delivery.","operatingModelImplications":"Recipes give a reproducible baseline that operators can validate with aicr validate.","sourceVerification":{"openedUrl":"https://developer.nvidia.com/blog/making-kubernetes-ai-ready-for-nvidia-vera-rubin-on-day-one/","referenceExcerpt":"It’s an early-adopter path that does not yet carry the production support and lifecycle qualification of a Supported recipe","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}