H2O AI and VMware by Broadcom
H2O AI Hybrid Cloud, the self-managed deployment of H2O AI Cloud, is validated and supported on vSphere Kubernetes Service (VKS), the Kubernetes runtime built into VMware Cloud Foundation (VCF). Organizations that run VCF install it with Helm and run predictive, generative and agentic AI on infrastructure they already operate, with the same administrators, lifecycle management, security boundary and GPUs as the rest of their estate.
Partnership overview
What customers get
- AI where the data lives: predictive, generative and agentic AI on data that can't leave your data center.
- One platform to run: VMs, containers and AI under one operating model, one set of skills and one security boundary.
- GPUs that stay busy: self-service engines start, pause and release GPU capacity on demand.
- Validated and supported: deployed, tested and supported as a validated stack on VKS.
How it works
H2O AI Cloud installs with Helm on a VKS cluster provisioned from the vSphere Supervisor. It stores data on vSAN through the vSphere CSI driver and runs GPU work on NVIDIA vGPU or GPU passthrough, through a GPU-enabled VM Class. Data scientists start H2O Driverless AI, H2O-3 or Notebook Engine engines sized to the job, and H2O AI Engine Manager creates, pauses and reclaims them as needed.
Resources
- Technical white paper: H2O AI on vSphere Kubernetes Service on VMware Cloud Foundation (VMware, October 2026): reference architecture, Helm deployment and validation steps.