Kubernetes adoption shows no signs of slowing. A 2025 CNCF survey found that 80% of organizations are running Kubernetes in production, up from 66% in 2023. With so many teams depending on it, the pressure to run workloads efficiently and securely at scale is greater than ever.
Multi-tenancy provides a powerful framework for running and managing diverse workloads on shared infrastructure. It enables organizations to optimize resource utilization, reduce costs, and simplify operations, while still maintaining strong isolation and security boundaries when it is carefully designed and implemented.
To learn more about the benefits of multi-tenancy and the future of Kubernetes, VMblog spoke with Lukas Gentele, co-founder and CEO of vCluster Labs.
++
VMblog: vCluster has recently made three key announcements aimed at advancing Kubernetes tenancy. Can you walk us through what’s new?
Lukas Gentele: The advancements are all focused on extending the tenancy spectrum: Private Nodes, Auto Nodes, and vCluster Standalone. Each of these releases builds on the last to give platform teams more flexibility.
Private Nodes allow tenants to attach their own dedicated nodes directly to a virtual cluster. Auto Nodes extends that with dynamic autoscaling powered by Karpenter, so those isolated environments can expand or shrink on demand across any infrastructure. And vCluster Standalone removes the dependency on a host cluster entirely, letting teams bootstrap Kubernetes directly on bare metal or VMs with vCluster itself as the foundation.
Together, these updates make it possible for organizations to choose the tenancy model that best fits their needs at any given time, without changing their overall architecture.
VMblog: Let’s talk about Private Nodes, which seem to be the foundation of these capabilities. How do platform engineering teams benefit?
Gentele: Private Nodes were built for cases where shared infrastructure is not enough. In heavily regulated environments or for high-performance workloads like AI and ML training, teams need stronger isolation than namespaces or shared nodes can provide.
With Private Nodes, every tenant operates with dedicated nodes that is invisible to other tenants. That means full freedom to bring your own runtime, CNI, CSI, or GPU operator, while vCluster continues to manage the control plane in the host cluster. Platform teams no longer have to run completely separate clusters for these use cases. They get strict isolation without giving up operational simplicity.
VMblog: What are Auto Nodes, and how do they extend Private Nodes?
Gentele: Private Nodes gave us hard isolation, and with that, the next challenge was elasticity. Without autoscaling, tenants would still be stuck overprovisioning nodes or leaving expensive hardware idle.
Auto Nodes solves that by embedding Karpenter into each virtual cluster, so scaling decisions happen at the tenant level. If workloads need more compute, Auto Nodes provisions it on demand. When demand drops, unused nodes are released. With vCluster, this works for the first time across public cloud, private cloud, and bare metal instead of being limited to a single public cloud like AWS. The result is full isolation and dynamic scaling in one model, something you could not achieve before without stitching together multiple systems.
VMblog: What use cases stand out as the biggest beneficiaries of dynamic autoscaling?
Gentele: There are three that really stand out.
- AI and GPU-driven workloads where capacity needs fluctuate dramatically.
- CI and CD pipelines that need bursts of compute for test jobs but do not justify maintaining large static pools.
- Multi-tenant SaaS platforms where each tenant environment can scale independently without impacting others.
These are all scenarios where elasticity is crucial, but isolation cannot be compromised.
VMblog: Karpenter is becoming widely adopted in cloud-native environments. Can you explain what it does and why it matters?
Gentele: Karpenter is a workload-aware autoscaler that provisions nodes dynamically based on the needs of unschedulable pods. Instead of relying on static node groups, it creates the right node at the right time, which makes clusters more efficient.
In managed services like EKS, Karpenter is already widely used, but it is tied to AWS. What we have done with Auto Nodes is embed Karpenter inside each virtual cluster so it works anywhere, across cloud providers, private data centers, or bare metal. That is the big difference. You get the benefits of Karpenter without being locked into one vendor.
VMblog: With AI and GPU-driven workloads on the rise, how do you see infrastructure strategy evolving for platform engineering teams?
Gentele: GPUs are high-value and difficult to come by, which means they need to be used as efficiently as possible. Many organizations are realizing that simply adding more hardware is not sustainable, especially when workloads are bursty or unpredictable. The focus has to shift toward better utilization, stronger isolation, and more flexible scaling strategies. By combining approaches like Private Nodes and Auto Nodes, platform teams can ensure that GPU resources are dedicated when needed and scaled back when idle, creating a balance between performance, cost, and security.
VMblog: You’ve also recently introduced vCluster Standalone. What does it enable for organizations?
Gentele: Standalone removes the extra step of running a host cluster before deploying virtual clusters. Traditionally, you needed something like EKS, Rancher, or Talos underneath vCluster. That meant separate vendors, contracts, and upgrade paths.
With Standalone, vCluster itself becomes the foundation. You can bootstrap Kubernetes directly on VMs or bare metal, and then run any tenancy model, from lightweight shared nodes to fully isolated private nodes with autoscaling, on top. It simplifies operations and unifies the stack under a single vendor-supported solution.
VMblog: What’s next for Kubernetes Tenancy?
Gentele: Tenancy is moving toward greater flexibility. Rather than locking into one model, teams want to choose the right level of isolation and scalability for each workload. That can mean lightweight environments for development, strict isolation for regulated industries, or elastic scaling for AI and GPU-driven workloads.
The rise of AI is making this even more important. GPU infrastructure is expensive, and utilization has to be high while still ensuring strong security boundaries. The future of Kubernetes tenancy will be about balancing those priorities so organizations can run diverse workloads efficiently, securely, and at scale.
##
