VMware says VCF Private AI Services runs as Supervisor Services in VMware Cloud Foundation and supports model runtime, RAG data indexing, API access, and tool registry components across Supervisor architectures with and without NSX.
VMware has detailed how organizations can deploy VMware Cloud Foundation Private AI Services across Supervisor architectures with and without NSX.
In a VMware Cloud Foundation Blog post, VMware says VCF Private AI Services includes several components for building and operating private AI applications: Model Runtime, Agent Builder, retrieval-augmented generation data indexing, an API Gateway, and an MCP Tools Registry.
The company describes these services as part of a broader VMware Private AI Foundation with NVIDIA effort in VMware Cloud Foundation 9.0. In a separate VMware Cloud Foundation Blog post on Private AI Services in VCF 9.0, VMware says the services are deployed into the VCF Supervisor as Supervisor Services and include capabilities such as Model Store, Model Runtime, Data Indexing/Retrieval, and Agent Builder.
Broadcom also said in its VMware Explore 2025 announcement that VMware Private AI Services would become a standard component of VMware Cloud Foundation 9.0. Broadcom listed native AI services including GPU Monitoring, Model Store, Model Runtime, Agent Builder, Vector Database, and Data Indexing/Retrieval.
According to VMware, inference endpoints for VCF Private AI Services run on VMware Kubernetes Service worker virtual machines. VMware also says the RAG function connects to an external vector database, rather than treating indexing and retrieval as a standalone feature isolated from the wider application environment.
That architecture matters for infrastructure teams because it ties private AI application services to VCF Supervisor operations. VMware’s posts position the Supervisor as the deployment surface for these AI services, while VKS worker virtual machines provide the runtime location for inference endpoints.
The latest VMware Cloud Foundation Blog post focuses on navigating Supervisor architectures with and without NSX. VMware’s framing suggests that organizations evaluating Private AI Services need to understand how their existing Supervisor networking architecture affects deployment choices.
The announcements show VMware and Broadcom continuing to fold AI application infrastructure into VMware Cloud Foundation rather than treating it as a separate platform layer. Broadcom’s announcement describes VMware Cloud Foundation as becoming an “AI native” platform, while VMware’s technical posts explain the service-level components that support that direction.
For enterprises already operating VMware Cloud Foundation, the key point is not just the presence of model-serving features. VMware is also describing supporting services around model storage, API access, data indexing and retrieval, and tool discovery. Those elements are commonly needed when moving from isolated model testing to applications that need governed access to internal data and services.
VMware’s description of an external vector database connection for RAG also indicates that customers will still need to plan surrounding data infrastructure. Private AI Services can provide data indexing and retrieval capabilities, according to VMware and Broadcom, but VMware’s deployment description still references an external vector database as part of the architecture.
Broadcom’s VMware Explore 2025 announcement said VMware Private AI Services would be a standard component of VMware Cloud Foundation 9.0. VMware’s subsequent technical material provides more detail on how those services are packaged and where they run.
Taken together, the sources show a product strategy centered on integrating AI infrastructure services directly into VMware Cloud Foundation. The practical deployment details — including Supervisor Services, VKS worker virtual machines, RAG integration, and support for Supervisor architectures with and without NSX — are the parts infrastructure teams will need to evaluate before adoption.
VMware has detailed how organizations can deploy VMware Cloud Foundation Private AI Services across Supervisor architectures with and without NSX.
The company describes these services as part of a broader VMware Private AI Foundation with NVIDIA effort in VMware Cloud Foundation 9.0.
Broadcom also said in its VMware Explore 2025 announcement that VMware Private AI Services would become a standard component of VMware Cloud Foundation 9.0.
Continue reading
