From the NVIDIA edition of November 2, 2026
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
NVIDIA · AI infrastructure / Kubernetes platform software · Global
- Publisher
- NVIDIA Technical Blog
- Original publication
- 2026-11-02
- Source retrieved
- 2026-11-03
- Event date
- 2026-11-02
What happened
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.
Why it matters
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 and measured results
Vendor validation on early-access reference clusters; one bus-bandwidth gate reported. Preview status stated by NVIDIA.
Limitations and uncertainty
Pre-release platform and Preview recipes; MIG not validated; tested versions are not stated minimums.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-03; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- 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
- Discovery questions
- 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?
- Value hypothesis
- An early readiness review reduces surprises when Vera Rubin ships.
- Potential engagement
- Readiness review of a current GPU Kubernetes stack against AICR Vera Rubin recipes.
- Unsupported claims
- Production readiness
- Performance beyond the single reported gate
- MIG support
Pre-sales engineering
Role takeaway
- Fit
- Relevant for Kubernetes-based GPU platforms planning Vera Rubin NVL72.
- Architecture and integration
- 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.
- Proof of value
- Run aicr validate on the current cluster and log gaps
- Re-validate when Supported recipes ship
Delivery
Role takeaway
- Work
- Readiness assessment and staged upgrade plan for Vera Rubin.
- Dependencies
- NVIDIA Supported recipe release
- Hardware delivery
- Driver GA
- Ownership
- Platform engineering team.
- Skills and adoption
- Training on DRA, ComputeDomains and IMEX.
- Governance checkpoints
- Acceptance tied to Supported recipes
- DPU security review
- Acceptance criteria
- 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
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Upgrading from GB200 means new driver branch, ConnectX-9/BlueField-4 networking, NUMA/NIC affinity checks and DRA-managed IMEX.
Governance
Who approves, reviews and stays accountable for outcomes?
Preview recipes should not be accepted as production baselines in contracts.
Security and privacy
What data, permissions and controls need testing?
BlueField-4 DPUs replace BlueField-3; DPU security posture needs fresh review (BlueField DPU tampering research was covered in an earlier edition).
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Platform teams need Kubernetes DRA and IMEX skills.
Procurement
What should contracts, pricing and exit terms secure?
Buyers should tie acceptance to Supported, not Preview, software status and confirm driver branches at delivery.
Operating model
Which teams own the service once it runs?
Recipes give a reproducible baseline that operators can validate with aicr validate.
Publication history
- 2026-11-02NVIDIA · Issue 585 resources
Stable resource ID: nvidia-aicr-vera-rubin-nvl72-kubernetes-preview-recipes-20261102