Broadcom Inc. unveiled the platform at VMware Explore 2026, positioning it as a production-ready path for enterprises moving from AI experimentation to deployment while keeping data sovereignty, cost predictability and governance under their own control. Built on VMware Cloud Foundation 9, the platform supports more than 150 open source and commercial AI models, including Nemotron 3, Gemma 4, cotomi, Qwen and GLM 5.2, and runs across heterogeneous GPU and CPU clusters from major hardware vendors.

Addressing the cost of scaling enterprise AI

VMware Private AI Cloud targets three cost drivers Broadcom says enterprises face when scaling AI: hardware capital expenditure, operational complexity and token economics. VMware Cloud Foundation 9 lowers hardware costs through NVMe memory tiering and cluster-wide storage deduplication, while a new AI Factory layer automates deployment and day-two operations to speed up time to first model deployment.

“VMware Private AI Cloud is the inflection point where enterprise private cloud and private AI infrastructure stop operating as separate disciplines and become one, enabling production inference workloads and agentic AI with the data sovereignty, compliance posture, and cost predictability their business demands,” said Ram Velaga, President, Infrastructure Software Group, Broadcom.

Governing autonomous AI agents as enterprise identities

A key piece of the announcement is AgentMinder, a new control plane that treats autonomous AI agents as enterprise-grade identities, binding their authority to a specific mission, approved tools and authorised resources, with runtime policy enforcement and audit visibility. The platform also adds vDefend enhancements for zero-trust security around agentic AI workloads, including detection of unauthorised shadow AI usage and distributed virtual patching.

Security built around AI-accelerated threats

Broadcom says the platform’s security architecture is aligned to NIST CSF 2.0 and designed to defend against AI-accelerated threats through microsegmentation, automated non-disruptive updates and continuous compliance monitoring. Tanzu Platform agent foundations enforce a deny-by-default runtime, meaning AI agents have zero access to APIs, networks or the internet unless explicitly granted, with an isolated credential store shielding credentials from agents entirely.