VMware, now part of Broadcom, has launched its own "AI Factory" product, borrowing a name popularised by long-time partner Nvidia but building a version that also supports rival AMD's hardware. The move matters because it shows VMware broadening its AI infrastructure ambitions beyond its close historical ties to Nvidia, giving enterprises a way to run AI workloads on AMD's Instinct GPUs and ROCm software rather than being locked into Nvidia's ecosystem, while still keeping Nvidia as a core partner.

The new VMware AI Factory, announced at the VMware Explore conference, effectively rebrands and expands the 2023 "VMware Private AI Foundation with Nvidia" offering, automating hardware provisioning, software deployment and lifecycle management, and adding features such as secure model-sharing across tenants, hybrid on-prem/cloud deployment, cost-aware workload placement, and AI governance tools to keep agents within set boundaries. It requires VMware Cloud Foundation and runs on servers from Cisco, Dell, Lenovo and Supermicro; a separate AMD-based version uses AMD Instinct GPUs and the open ROCm ecosystem. Broadcom executive Prashanth Shenoy described it as an "evolution" of the Nvidia-based product that now supports multiple accelerator architectures, though the current AI Factory doesn't yet formally align with Nvidia's own AI Factory reference architecture, and VMware hasn't ruled out doing so in future.