Broadcom Unveils VMware AI Factory With 'AgentMinder' Identity Governance for AI Agents

Broadcom introduced VMware AI Factory at VMware Explore 2026 in Las Vegas on August 31, 2026 — a software-defined foundation within VMware Private AI Cloud designed to take enterprises from bare-metal infrastructure through to live model inference, combining VMware Cloud Foundation's automation with Tanzu Platform, vDefend, Avi Load Balancer, and new AI governance capabilities.

AgentMinder: Treating AI Agents as Identities

The most notable component is AgentMinder, which treats AI agents as enterprise-grade identities in their own right — binding each agent's authority to a specific mission, a defined set of approved tools, and authorized resources, with runtime policy enforcement to ensure least-privileged access for every tool invocation. It also delivers compliance-grade auditability, giving security teams a record of what an agent was authorized to do and what it actually did.

The Infrastructure Speed Claim

• Broadcom says VMware Cloud Foundation's automation can cut the time from bare-metal server deployment to serving a first AI model from several weeks down to hours, by automating hardware provisioning, software stack enablement, and full lifecycle operations
• VMware Private AI Cloud supports more than 150 open source and commercial AI models, with multi-tenant model sharing through isolated namespaces
• An AI Gateway component is meant to provide unified model governance spanning both on-premises and cloud environments, rather than requiring separate governance tooling for each

What This Means If You're Deploying AI Agents Internally

AgentMinder's identity-based approach to agent governance addresses a real, growing gap: most organizations still lack a systematic way to scope what an AI agent is actually allowed to do, versus what it technically can do given the credentials it holds. For any enterprise running or planning to run autonomous agents against production systems, treating agent permissions with the same rigor as human identity and access management — least privilege, auditability, mission-scoped authority — is the practical takeaway here, independent of which specific vendor's tooling you use to get there.