Guest: Saiyam Pathak (LinkedIn)
Company: vCluster Labs
Show Name: KubeStruck
Topic: Kubernetes, Cloud Native

AI infrastructure is moving back on-premises, and vCluster just made that transition seamless. At KubeCon North America, vCluster announced a strategic partnership with NVIDIA to bring Kubernetes orchestration to DGX GPU machines. This isn’t just another vendor partnership. It’s a reference architecture designed for organizations that need to run AI workloads locally while keeping sensitive data secure.

The announcement addresses a critical enterprise pain point. Organizations want to leverage large language models and AI capabilities, but they don’t want their proprietary data, internal documents, or code leaving their infrastructure. “Everyone wants their data to remain local. Everyone wants that data to be secure, and  they want to inference their data while keeping their internal documents and code within their own infrastructure,” explains Saiyam Pathak, Head of Developer Relations at vCluster.

NVIDIA’s DGX platform provides the GPU supercomputing power. vCluster brings the Kubernetes layer that makes those resources accessible and manageable. The result is what Pathak calls “an out-of-the-box Kubernetes experience” on GPU infrastructure that enterprises can deploy in their own data centers.

The partnership includes reference architectures that combine three key vCluster capabilities: private nodes, auto nodes, and vNode. These components work together to deliver complete isolation for AI workloads. This isolation is critical when different teams or projects need to share expensive GPU resources without risking data leakage or security vulnerabilities.

The timing is strategic. AI infrastructure has emerged as one of the biggest technology investments for 2024 and beyond. Enterprises are moving past experimentation with cloud-based AI services and building their own infrastructure stacks. They need GPU power to train and run models. They need Kubernetes to orchestrate workloads efficiently. And they need proven architectures that won’t require months of trial and error.

vCluster’s approach leverages over a year of development work on specific components that now come together in this NVIDIA partnership. The reference architectures mean organizations don’t have to start from scratch. They can deploy proven patterns for running Kubernetes on DGX infrastructure and focus on their AI applications rather than infrastructure plumbing.

This partnership signals a broader shift in the AI infrastructure market. While cloud providers offer powerful AI services, enterprises with sensitive data or specific compliance requirements are investing in on-premises GPU infrastructure. They want the flexibility of Kubernetes, the power of NVIDIA GPUs, and the control that comes with running everything in their own environment.

For platform engineering teams tasked with building AI infrastructure, this partnership offers a clear path forward. The combination of NVIDIA’s hardware leadership and vCluster’s Kubernetes expertise provides a foundation for secure, scalable AI workloads that stay within organizational boundaries.