.jpeg)
Table of Contents
This year’s KubeCon in Atlanta was one for the books. For vCluster Labs, it was not just about the energy on the show floor but the redefinition of how organizations will run AI workloads on Kubernetes.
At Booth #421, we officially launched the Infrastructure Tenancy Platform for NVIDIA DGX, a Kubernetes-native platform designed to maximize GPU efficiency across NVIDIA-powered environments -> from private AI supercomputers to hyperscalers and neoclouds.
The response was overwhelming: packed demo sessions, a standing-room-only happy hour, and endless conversations about the future of tenancy, GPUs, and AI on Kubernetes.
Major Announcement
A new Reference Architecture for NVIDIA DGX systems is now available, offering architectural guidance for building secure, scalable Kubernetes environments optimized for NVIDIA AI infrastructure. This announcement provides a clear vision, to make AI infrastructure dynamic and efficient ensuring security, flexibility, and speed.
vCluster’s most recent innovations give enterprises a single composable layer for GPU-intensive workloads:
- vCluster Private Nodes & Auto Nodes: Create the Kubernetes cluster with hosted controlplane on the host cluster and nodes joined privately. Enable dynamic autoscaling of GPU and CPU capacity using Karpenter automation across clouds and bare metal.
- vCluster VPN: Secure hybrid connectivity between control planes and nodes with Tailscale integration for burst-to-cloud scenarios.
- NVIDIA Base Command Manager Integration: Seamless lifecycle and scaling management for DGX systems.
- KubeVirt Integration: Partition large bare-metal systems into smaller, isolated compute units.
- Netris Integration: Network isolation for each tenant with full lifecycle automation.
- vNode Runtime: Rootless, secure sandbox for multi-tenant workloads on shared clusters.
Together, these features deliver cloud-like elasticity and governance for workloads in the cloud and on private AI infrastructures.
From a review on The New Stack: “This new Kubernetes distribution provides flexible, multi-tenant GPU orchestration with dynamic scaling and isolation.”
From a review on AInvest: “The platform offers a unified framework for deploying and managing AI workloads on AI supercomputers in the private cloud and on top of hyperscalers and emerging neoclouds.“
KubeCon Sessions & Keynotes
vCluster had a strong presence across multiple sessions:
- Nov 10 - Keynote: AI-Ready Platforms: Scaling Teams Without Scaling Costs, Saiyam Pathak discussed how virtual clusters and autoscaling reshape how AI teams scale infrastructure.
- Nov 10 - AI Clusters Anywhere, Lukas Gentele presented Autoscaling GPU Clusters on Hyperscalers, Neoclouds & Bare Metal.
- Nov 12 - Open Source at the Edge: Hardware, Firmware & AI Stacks, Saiyam and Miley Fu where they showcased EchoKit device and how to create AI enabled voice assistant.
- Nov 12 - Building Resilient Cloud Native Infrastructure in the Second Decade - TAG Operational Resilience, Rafael Brito, StormForge; Mario Fahlandt, Kubermatic; Saiyam Pathak, vCluster; Carolina Valencia, KrolCloud; Nabarun Pal, Broadcom where they discussed the TAG operational resilience, initiatives and sub projects like Green Reviews.
- Nov 13 - Help! My LLM is a Resource Hog: How We Tamed Inference with Kubernetes and Open Source Muscle, Hrittik and Aditya where they discussed LLM's, performance, GPU sharing and tenancy using vCluster.
- Demos: Multiple demos were given on: GPU Sharing using MIG, GPU & CPU Private AutoNodes on EKS and Internal Developer Platforms with vCluster and API-Enabled Platforms on Kubernetes got the booth buzzing with learning energy.
- Book Signing: We also held multiple book signings throughout the event to distribute free copies of the book and appreciate the authors for their work. Alongside that our very own Saiyam Pathak did his first book signing
vCluster Evening Event: The Future of AI and Kubernetes – with NVIDIA and JPMorganChase
One of the most talked-about events at KubeCon was the “Future of AI and Kubernetes” Happy Hour & Fireside Chat hosted at Top Draft, Omni Atlanta.
The evening was packed with a full house of platform engineers, AI architects, and open-source contributors. Fabian(CTO vCluster) joined speakers from NVIDIA and JPMorgan Chase, sparking a deep discussion around tenancy maturity models, what is happening in AI infrastructure, and the future of AI and Kubernetes.
Attendees shared how they’re moving from namespace-based tenancy to vCluster-powered isolation, leveraging features like Private Nodes, Karpenter-based scaling, and Standalone clusters that eliminate dependency on the host Kubernetes cluster.
The energy in the room said it all: this is where the AI + Kubernetes conversation is heading next.
What We Heard from the Community
Throughout the week, one theme echoed across the booth: “vCluster is solving real problems.”
- Platform teams loved how vCluster brings tenancy flexibility, from shared clusters to hosted control planes and standalone setups.
- Many attendees shared plans to migrate from namespace-based tenancy to vCluster isolation for improved governance and resource fairness.
- Bare-metal operators were excited about how vCluster enables AI workloads on NVIDIA GPUs without sacrificing elasticity.
- We met existing open-source users, customers, and new partners -> from hyperscaler teams to enterprises running private GPU clouds.
The excitement was amazing, and the booth stayed busy from open to close every single day.
Why This Announcement Matters
vCluster as a center for AI infrastructure tenancy design, is a signal that the future of Kubernetes is AI-driven, multi-tenant, and hybrid by design.
As enterprises race to scale AI workloads, they need efficient GPU orchestration and strict tenant isolation. vCluster delivers that with the elasticity of the public cloud and control of private infrastructure.
A BIG “Thank You” to our Community
From our speakers, to authors, and to every attendee who stopped by the vCluster Booth, thank you for being part of this incredible week. We couldn’t have asked for a better community moment:
- Packed sessions
- Packed fireside chat
- Excited partners
- Engaged customers
- Endless hallway conversations about what’s next for AI and Kubernetes
Learn More
- Read the full announcement: vCluster Press Release
- Our new free ebook: GPU-Enabled Platforms on Kubernetes
- Follow @vClusterLabs for updates from the community.
Until Next Time
KubeCon NA 2025 was a celebration of what’s possible when innovation meets community. The future of Kubernetes multi-tenancy is already here and it’s powered by vCluster.
