Guest: Lukas Gentele (LinkedIn)
Company: vCluster Labs
Show Name: KubeStruck
Topics: Kubernetes, Cloud Native

AI has pushed Kubernetes into new territory — one where workloads are data-heavy, GPU-intensive, and often privacy-sensitive. In this evolving space, vCluster Labs is rethinking how infrastructure teams can manage GPU capacity across business units while maintaining fairness and control.

Kubernetes Adapts to AI

As Lukas Gentele, Founder and CEO of vCluster Labs, explains, Kubernetes itself is evolving to support specialized hardware. “Initially, it was built for long-running web apps,” he says, “but now, with advancements like Dynamic Resource Allocation, you can specify exactly what kind of GPU you need — even distinguishing between commodity and next-gen GPUs like NVIDIA’s Blackwell series.”

This built-in flexibility means any innovation added to Kubernetes core immediately benefits vCluster users. But vCluster goes further, helping enterprises manage resource fairness and efficiency at scale.

Fair Use and Flexible GPU Allocation

Running AI workloads on-premises or in private clouds introduces a new challenge: how to share GPUs among teams securely. “Let’s say a company has 1,000 GPUs across multiple business units,” Gentele explains. “You don’t want to statically assign them. Auto Nodes lets you pool GPUs and assign them dynamically based on need.”

This elasticity allows one team to borrow GPUs for training while another team uses them for inference — all managed automatically by vCluster’s Auto Nodes system, which integrates Karpenter for intelligent scheduling. The result is higher utilization, reduced waste, and equitable access across departments.

Bridging AI and Cloud-Native Principles

Beyond optimization, vCluster’s approach aligns with cloud-native philosophy — automation, portability, and scalability. It enables enterprises to treat GPUs like any other elastic resource, freeing them from rigid hardware allocations. For organizations handling sensitive data or regulated workloads, it also ensures that training happens within controlled environments — without losing the flexibility of cloud-native scaling.

Takeaway

AI workloads are redefining what “multi-tenancy” means in Kubernetes. With vCluster Auto Nodes and support for DRA, vCluster Labs is giving enterprises the tools to scale GPU usage intelligently — balancing performance, privacy, and cost in ways traditional cluster setups can’t.