Broadcom has announced the VMware AI Factory, a software-defined private AI platform designed to accelerate the transition from bare-metal servers to production-ready AI models. This platform enhances governance, infrastructure automation, and workload isolation.
Unveiled during VMware Explore 2026, the VMware AI Factory serves as the foundation for VMware’s Private AI Cloud. It combines VMware Cloud Foundation (VCF), validated server hardware, GPU infrastructure, model services, and lifecycle management into a cohesive environment for deploying enterprise AI workloads that handle sensitive data.
Broadcom claims that this platform can reduce the time required to deploy infrastructure and serve an initial AI model from weeks to hours. Its automation features address operational bottlenecks that enterprises encounter when scaling GPU-intensive AI applications, including hardware provisioning, software stack enablement, and Day 2 lifecycle operations.
A key cybersecurity feature of the VMware AI Factory is the introduction of secure AI sandboxes and governance controls. Broadcom stated that these virtualized container spaces will isolate dynamically generated and executed code by AI agents.
This security control layer is designed to govern how agents are invoked, determine the tools they can access, and validate their outputs before those outputs trigger actions.
This helps organizations mitigate the potential fallout from failures of autonomous agents, unsafe tool use, prompt-driven behaviors, or malicious code execution in enterprise AI workflows.
This approach is especially relevant for retrieval-augmented generation (RAG), workflow automation, coding agents, and other AI applications that connect models to internal data sources, APIs, or infrastructure management functions.
VMware AI Factory Features
| Feature | Technical capability | Enterprise security and operational impact |
|---|---|---|
| Secure AI sandboxes | Virtualized containers isolate agent-generated code execution | Helps contain untrusted or dynamically generated workloads |
| Agent governance layer | Controls agent invocation, permitted tools, and output validation | Reduces risk from excessive agent permissions and unsafe actions |
| Multi-tenant model sharing | Isolated namespaces enable secure model sharing between tenants or business units | Preserves data privacy while avoiding duplicate model deployments |
| AI Gateway | Provides a common interface for local and cloud-hosted models | Adds prompt routing, application authorization, token controls, and rate limiting |
| Unified model gallery | Deploys and manages inference and RAG workflows across VMs, containers, and GPUs | Centralizes model governance and operational oversight |
| Observability | Tracks token throughput, latency, compute usage, and memory utilization | Supports cost management, capacity planning, and performance monitoring |
| Bare-metal automation | MetalSoft integration provisions or repaves heterogeneous servers through VCF | Eliminates separate vendor tooling and speeds hardware lifecycle operations |
Broadcom emphasizes AI tokenomics, the cost associated with generating and consuming AI tokens, as a crucial concern for enterprises. The VMware AI Factory allows teams to pool GPU resources, enabling multiple models to share hardware rather than each requiring a dedicated infrastructure for their workload.
The platform provides visibility into token throughput, latency, compute consumption, and memory utilization. This insight allows IT, security, and data science teams to manage model usage, identify inefficient workloads, and implement policy controls to prevent unexpected AI-related costs from escalating.
The AI Gateway also introduces centralized consumption controls for models deployed both on-premises and in cloud environments, with capabilities planned for intelligent prompt routing, token and usage rate limits, and application authorization.
VMware AI Factory is compatible with certified Dell PowerEdge systems and VCF AI ReadyNodes from Cisco, Lenovo, Supermicro, and other vendors. Broadcom is also collaborating with AMD to integrate VCF with AMD Instinct MI350-series GPUs and the ROCm software ecosystem.
Customers will be able to run over 150 open-source and commercial models, including Nemotron 3, Gemma 4, Cotomi, Qwen, and GLM 5.2. The platform aims to offer enterprises a controlled private cloud alternative for deploying models while maintaining data sovereignty, infrastructure choice, and governance over increasingly autonomous AI workloads.
Prevent incidents due to slow investigations. Power your Tier 1 with threat intelligence from 15K SOCs: Integrate TI Lookup in your SOC
Attackers abused compromised ConnectWise Cloud Backup credentials to access Microsoft 365 customer environments, prompting ConnectWise…
Anthropic says an internal Claude research model reconstructed how the National Security Agency generated seeds…
Two WordPress Core vulnerabilities can be chained to execute remote code without authentication, according to…
A newly disclosed OpenBSD kernel exploitation chain combines a shared memory accounting flaw with a…
GhostAction attackers have expanded their GitHub credential theft campaign with a second stage that disguises…
Indian authorities have warned that SideCopy hackers are targeting Indian organizations, including academic institutions, with…