Broadcom launches VMware Private AI Cloud for enterprises

Broadcom has introduced VMware Private AI Cloud, built on VMware Cloud Foundation 9.

The platform is intended to let enterprises build, run and govern artificial intelligence workloads where their data is held, while also running traditional applications on the same private cloud setup.

Broadcom is pitching the offering as many large organisations move from testing AI tools to putting them into production. That shift has pushed issues such as data sovereignty, governance and cost control higher up the agenda for technology buyers.

According to Broadcom, VMware Private AI Cloud supports more than 150 open source and commercial AI models. It is designed to run inference workloads, agentic applications and other enterprise workloads across mixed GPU and CPU clusters.

VMware Cloud Foundation 9 lowers hardware costs through NVMe memory tiering and cluster-wide storage deduplication, Broadcom said. It also supports GPUs, CPUs and accelerators from several vendors, alongside server hardware from major original equipment manufacturers and original design manufacturers.

Broadcom has also launched VMware AI Factory as part of the wider push around the new platform. It described the software as the foundation of VMware Private AI Cloud, adding automation for deploying AI-ready infrastructure and handling ongoing operations.

Operational features include token monitoring, multi-tenant model sharing, GPU and virtual GPU tracking, and an AI metrics observability dashboard. Broadcom said these functions are intended to help customers manage resource use and the economics of running AI services.

Validated AI models for VMware Cloud Foundation include Nemotron 3, Gemma 4, cotomi, Qwen and GLM 5.2. Broadcom said customers will be able to run them through built-in services on the platform.

Security focus

Security is a central part of the launch, with Broadcom linking the product to a defence-in-depth approach aligned with NIST CSF 2.0. Measures include automated updates, microsegmentation, virtual patching, and web application and API protection.

New features in VMware vDefend are intended to address risks tied to agentic AI workloads. Broadcom said the software will identify agentic AI components by monitoring traffic flows, detect unauthorised use of so-called shadow AI, and apply distributed virtual patching through AI-generated intrusion detection and prevention signatures.

Broadcom also outlined updates to TrueSource by Broadcom, which it said provides verifiably built open source software. The work covers parts of the Java ecosystem, Python and Node.js, as well as data services including PostgreSQL, RabbitMQ, MySQL and Valkey.

Agent controls

Another part of the announcement is AgentMinder, a new control plane for autonomous AI agents. Broadcom said the system treats agents as enterprise identities and ties their authority to specific missions, approved tools and authorised resources.

AgentMinder adds runtime policy enforcement and audit trails so companies can track what agents are doing and limit access at the point of each tool invocation, Broadcom said. The controls are aimed at enterprises concerned that autonomous software agents may exceed their intended remit or gain access to sensitive systems.

Broadcom also detailed new AI-ready data foundations in Tanzu Platform. The tools are intended to help data owners build pipelines across structured and unstructured information and publish the resulting data products as governed services for developers and agents.

According to Broadcom, Tanzu Platform has also been updated with agent foundations that enforce a deny-by-default runtime. Under that model, agents do not gain access to APIs, networks, MCP servers or the internet unless access is expressly granted, while credentials are held in an isolated store that agents cannot directly view.

Ram Velaga, President, Infrastructure Software Group, Broadcom, set out the company's position on how private cloud and AI infrastructure are converging.

"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 Velaga.