Broadcom Launches VMware AI Factory for Private Cloud AI

Broadcom has introduced VMware AI Factory, a software-defined foundation designed to accelerate enterprise AI deployments in private cloud environments. The platform, announced at VMware Explore 2026 in Las Vegas, addresses a persistent challenge: the complexity and time required to move from physical infrastructure to operational AI models.

According to Paul Turner, chief product officer of Broadcom's VMware Cloud Foundation Division, enterprises struggle with the journey from "metal to model" — a process that is typically slow and expensive. VMware AI Factory aims to compress deployment timelines from weeks to hours through automated infrastructure provisioning and unified lifecycle management.

Why it matters

As enterprises race to deploy AI capabilities, infrastructure bottlenecks have become a critical constraint. The ability to run AI models where data already resides — without cloud migration costs or latency — represents a significant operational advantage. By automating bare metal provisioning and enabling GPU resource pooling, VMware AI Factory addresses both the speed and economics of private AI deployment, potentially shifting how organizations evaluate build-versus-buy decisions for AI infrastructure.

Automation and resource efficiency

The platform's core value proposition centers on automation. VMware AI Factory handles hardware provisioning, software stack enablement, and end-to-end lifecycle management as integrated operations rather than separate workflows. This consolidation eliminates the need for vendor-specific tools across heterogeneous hardware environments.

A key economic feature is GPU resource pooling. Rather than dedicating infrastructure to individual workloads, VMware Cloud Foundation allows organizations to share GPU resources across multiple teams and models. The platform includes a unified model gallery that provides a single interface for deploying and managing model inference and retrieval-augmented generation (RAG) workflows, with built-in observability for token throughput, latency, and resource utilization.

Hardware partnerships and vendor flexibility

Broadcom is positioning VMware AI Factory as hardware-agnostic, supporting certified systems from Dell, Cisco, Lenovo, and Supermicro. A notable collaboration with AMD will integrate AMD Instinct GPUs and the ROCm software ecosystem, offering an alternative to NVIDIA-centric architectures.

The platform also includes a partnership with MetalSoft to deliver bare metal automation that reduces physical server provisioning time from weeks to minutes. This integration extends VMware Cloud Foundation's operational workflows to the physical layer, allowing IT teams to provision or repave servers from multiple vendors through a single management console.

Model support and new services

VMware AI Factory customers can access more than 150 open source and commercial AI models, including Nemotron 3, Gemma 4, Qwen, and GLM 5.2. Forthcoming features include multi-tenant model sharing through isolated namespaces, an AI Gateway for unified model governance across on-premises and cloud environments, and secure AI sandboxes for isolating agent-generated code execution.

The platform's focus on "AI tokenomics" — the cost structure of token-based AI consumption — reflects growing enterprise concern about unpredictable cloud AI expenses. By enabling private deployment with shared infrastructure, VMware AI Factory offers a path to more predictable cost models.

These details were first reported by Broadcom in a press release issued August 31, 2026.