Piotr Zaniewski: vCluster: GPUs on demand

Why this conversation

The host had just seen Piotr Zaniewski’s talk in event hall 2, and started from it: “So AI disrupting all the industry and you just show me something incredible.” Piotr Zaniewski is Head of Engineering Enablement at vCluster Labs. The host recorded the conversation on 2 November 2026 at SREday Amsterdam.

Piotr’s answer starts from the same disruption, and from the people it can leave out: “we have to be able to provide for our developers, make sure that they don’t stay behind.” What they need now is hardware: “Developers want to test with GPUs, want to provision them.” And it falls to the platform team: “We want to provision GPUs for them as platform engineers, help them.” That, Piotr said, “is one of the areas vCluster can help with.”

From virtual Kubernetes to AI infrastructure

The host still associated vCluster with Kubernetes, and Piotr agreed that is where it began: “So vCluster started as a virtualized Kubernetes, which is really a virtualized Kubernetes control plane component inside of a pod and some networking on top.” From there it grew: “However, we increasingly started supporting AI infrastructure.”

The use case is a provider that has to hand out a large amount of bare metal for GPUs to the customers waiting for it. “This is where our product suite can help with slicing and dicing bare metal servers, virtual machines, whether it’s in Kubernetes or Slurm.” Around the hardware come the usual questions of sharing it: “And we can help with tenancy, isolation, and all the related concerns.”

Cluster API, Argo CD and the platform engineers

The host asked how it connects with Cluster API. “Right, so Cluster API is a Kubernetes provider, so it can create Kubernetes clusters.” vCluster works with it and with other projects: “We integrate within vCluster with several projects such as Argo CD, Cluster API, and others”. They help provision vClusters and regular clusters and manage workloads, so that cluster management and isolation stay “under control of platform engineers”. In Piotr’s words, Cluster API and the others “are provision mechanisms that you can integrate from within your virtual cluster.”

Getting started, for free

“So, the best way go to vcluster.com or vcluster.com/docs.” There are blogs, a YouTube channel, and people from vCluster at KubeCon and at CNCF events like this one. Much of it runs on your own machine: “There’s a lot of things you can install directly on your computer.” Try it, and move to the wider product range when you are ready.

Piotr added one recent change: “recently we have significantly revamped our free and open source tier.” There are now two tiers that cost nothing: “So we have two tiers, vCluster open source and also the free tier where you can use our products with almost all features on for free.” That suits small companies, Piotr said, “or, you know, home lab enthusiasts like me.”

When it comes out

The episode is edited and in the release order. Like every episode it goes out on YouTube first, then on Apple Podcasts, Spotify and Amazon Music.