VMware AI Factory: Broadcom Opens AI Infrastructure to Multiple Hardware Vendors

Companies that want to run AI in their own data centers rather than relying on external cloud services quickly face an infrastructure decision: which hardware, which software, which vendor — and how to avoid long-term dependencies? Broadcom provided a direct answer on August 31, 2026, at VMware Explore in Las Vegas.

With VMware AI Factory, Broadcom has introduced the software-defined foundation for a private AI cloud. The key: the architecture supports certified hardware from multiple vendors simultaneously — Cisco, Dell Technologies, Lenovo, and Supermicro provide the so-called AI ReadyNodes — and supports both NVIDIA GPUs and AMD Instinct GPUs with the open ROCm ecosystem. Tying a private AI stack to a single hardware vendor is now a thing of the past.

What VMware AI Factory Delivers in Practice

At its core, VMware AI Factory combines VMware Cloud Foundation (VCF) with automated infrastructure provisioning and lifecycle management. Organizations can run AI integration and automation on their own hardware without switching between different tools for each step — from hardware provisioning to model operations.

• Infrastructure automation: From bare-metal provisioning to first model deployment in hours instead of weeks
• GPU resource pooling: Multiple teams share GPU capacity on a unified infrastructure without redundant setups
• Model Gallery: Single interface for deploying and managing over 150 models — including Gemma 4, Qwen 3.7-Max, and Nemotron 3
• AI Gateway: Unified governance with intelligent prompt routing and token rate limiting
• Secure AI sandboxes: Isolated execution environments for agent-generated code with validation controls

Hardware Freedom Instead of Vendor Lock-In

One of the most concrete announcements is the new partnership with MetalSoft. The integration enables heterogeneous bare metal automation directly through the VCF management console — provisioning, configuring, and integrating physical servers from different vendors: in minutes instead of weeks. For organizations wanting to scale AI workloads on-premise, this is a meaningful step forward. Infrastructure flexibility increases without a proportional rise in operational complexity.

Broadcom highlights control over AI tokenomics as a core benefit — meaning cost control and resource allocation when running large language models in-house. Token consumption, throughput, latency, and GPU utilization are measured uniformly and can be managed through the AI Gateway. For organizations that want to not only test AI but also allocate costs internally and budget for it, this is an important operational capability.

What This Means for Mid-Sized Companies

VMware AI Factory is primarily an enterprise platform — getting started requires existing VMware Cloud Foundation licenses. For mid-sized companies, the announcement is still relevant in several respects.

• Supply chain price pressure: As enterprise customers become less tied to individual GPU vendors, competition among hardware suppliers increases — mid-sized companies benefit indirectly through lower cloud prices over time.
• Signal for architecture decisions: Multi-vendor AI infrastructure is becoming the standard. Companies planning internal AI infrastructure today should make vendor neutrality an explicit requirement — not wait until the first contract is signed.
• Open models are becoming enterprise-ready: Over 150 supported models on one platform shows that open-weight models are increasingly production-ready — a solid foundation for on-premise AI projects in mid-sized environments as well.
• Internal AI cost management: The tokenomics principle scales down — SMEs can benefit from similar governance approaches once AI tools are used across multiple departments.

What to Assess Now

If your organization currently sources AI workloads exclusively through external services, a structured review is worthwhile: which data cannot leave your premises? How do cloud costs develop as AI usage grows? Which internal processes are suited for an on-premise AI solution? Answering these questions systematically is part of strategic IT consulting before infrastructure decisions are made. What NoviCogi has learned from working with mid-sized clients is documented in the references and cases.

Background on structural bottlenecks in AI infrastructure — from chip shortages to data center capacity — is covered in our article The New AI Bottleneck.

This article was created with AI assistance and editorially reviewed.