Broadcom puts 150+ models in VMware private cloud

Broadcom launched VMware Private AI Cloud with an on-premises Model as a Service that supports more than 150 open-source and commercial models. Broadcom’s announcement frames the product as a private path for inference, agentic applications, and conventional workloads, but the model count is a catalog claim rather than an independent performance result. For operators deciding between hosted and private inference, Nvidia’s recent capex analysis is the useful companion: a larger model shelf does not remove GPU utilization, power, or procurement constraints.