- H2O AI Cloud is a validated solution for VMware vSphere Kubernetes Service in VMware Cloud Foundation, announced by Broadcom on November 9, 2026.
- H2O AI Hybrid Cloud, which you run yourself, runs on VKS, so your data, models, and GPUs stay in your data center.
- Your platform team starts from a documented deployment. Broadcom's reference architecture tested sign-in, GPU-backed AutoML, model deployment, an AI app, and a RAG assistant with an agent run.
- The validation is functional. Broadcom's reference architecture reports no performance or sizing results.
- Air-gapped installs are covered by mirroring container images to a local registry.
If your private cloud runs on VMware Cloud Foundation, you can now run H2O AI Cloud on it. Your data, models, and GPUs stay in your own data center. H2O AI Cloud is now a validated solution for VMware vSphere Kubernetes Service (VKS), the Kubernetes runtime in VMware Cloud Foundation (VCF). That means VCF customers can run H2O.ai's predictive, generative, and agentic AI on the private cloud they already operate. Broadcom announced the validation on November 9, 2026, at KubeCon North America.
H2O AI Cloud runs on VKS inside your data center, and Broadcom tested five checks on it.
What to look for: everything sits inside the dashed line. Yellow is H2O.ai software. Grey is your VMware infrastructure. The black bar lists what Broadcom tested, not how fast it ran.
What do you get from running H2O AI Cloud on VKS?
Many organizations cannot send sensitive data to someone else's cloud. Their data and GPUs already sit in a VMware private cloud. A separate AI stack means new infrastructure, new skills, and a new security review.
H2O AI Cloud on VKS avoids that. AI runs as a workload on the Kubernetes platform your team already operates. Bring the AI to your data, not your data to the AI.
- Platform and infrastructure teams: run AI on VKS clusters in VMware Cloud Foundation, on the operating model you already know, instead of a separate AI stack.
- AI and data leaders: use one platform for predictive, generative, and agentic AI. Broadcom tested the build, deploy, app, and assistant steps on a single cluster.
- Security and compliance teams: keep data, models, and GPUs in your own data center, under your own controls. For an air-gapped install, you mirror images to a local registry.
- Business owners: put AI to work on data you cannot send outside, such as scoring risk, answering questions from your documents with citations, and automating work with agents.
That is sovereign AI in practice: you own your agents, models, data, and results, and you decide where they run.
What did Broadcom announce about H2O AI Cloud?
On November 9, 2026, at KubeCon North America in Salt Lake City, Broadcom named H2O AI Cloud a validated solution for VKS. The VMware Cloud Foundation division shared the news in a blog post by Ahmar Mohammad. It lists H2O AI Cloud under AI, machine learning, and advanced analytics.
The same day, Broadcom said more than 20 new partner solutions were validated for VKS. Its press release names H2O AI Cloud among them. Broadcom's Kubernetes Ecosystem Partners page also lists H2O AI Cloud and links to the reference architecture.
Tom Stenger, VP of Partnerships at H2O.ai, said in Broadcom's blog: "H2O AI Cloud on VKS lets enterprises run AI where their data lives, on the VMware infrastructure they already trust."
What is H2O AI Cloud?
The H2O Sovereign AI Platform is delivered as H2O AI Cloud. H2O AI Super Agent orchestrates predictive, generative, and agentic AI in one system, with observability built in.
You choose who runs it. H2O AI Managed Cloud is run by H2O.ai. H2O AI Cloud for Government is run by H2O.ai for the US government. H2O AI Hybrid Cloud is the form you run yourself, in your own cloud, on-premises, or fully air-gapped. The VKS validation applies to the form you run yourself.
What did Broadcom test on VKS?
In October 2026, Broadcom published a reference architecture for H2O AI Cloud on VKS. It describes the deployment and the checks the platform had to pass. The cluster used NVIDIA GPU nodes for the GPU-backed checks.
| Check | H2O.ai capability |
|---|---|
| Sign in through OIDC | H2O AI Cloud |
| Run an experiment on a GPU-backed engine | Driverless AI |
| Deploy a model and score it | H2O MLOps |
| Launch an app on its own address | H2O AI App Store |
| Answer a question with citations, then run an agent | Enterprise h2oGPTe, using retrieval-augmented generation |
Read the results the right way. The test cluster was small: one control plane node and three workers, one of them with GPUs. The training data was the public Iris dataset. This is a functional validation. It shows how the platform deploys and works on VKS. It does not report performance, scale, or sizing, so size your own cluster against your own workloads.
Broadcom also describes air-gapped installs. You mirror the container images to a local registry.
What this means for owning your AI
H2O.ai created the trusted Sovereign AI Platform because organizations should own their AI: their agents, models, data, and results, run anywhere from cloud to air-gapped.
Running anywhere is a promise about choice. VKS adds a place many enterprises already trust. Your infrastructure stays your decision.
Broadcom brings the infrastructure layer. H2O.ai brings the AI platform and the expertise to reach production, from top-ranked data scientists, including Kaggle Grandmasters, to forward-deployed engineers embedded with customer teams.
Who is this for?
It fits organizations that already run VMware Cloud Foundation and want private AI. If you want to keep data, models, and GPUs in your own data center, and run AI on the operating model your team already knows, start here.
How do you get started?
Read Broadcom's reference architecture. It documents the installation and points to a companion repository with manifests and Helm values. Then pick a first use case, such as risk scoring or a document assistant with citations.
Broadcom resources
- Broadcom blog: the VKS validations, including Tom Stenger's quote.
- Broadcom press release, November 9, 2026, on the VM API contribution and the new validated solutions.
- Broadcom Kubernetes Ecosystem Partners page, listing H2O AI Cloud.
- Broadcom reference architecture for H2O AI Cloud on VKS, October 2026.
:quality(75))