Broadcom, which owns VMware, told attendees at the VMware Explore conference a few weeks ago that VMware Cloud Foundation has been made AI native. The announcement positions the platform to host machine learning workloads more directly inside customer environments, a move that follows broad industry momentum around large language models and other generative systems.
The timing of that statement matters. Broadcom still faces criticism over changes to VMware’s licensing policies since completing the takeover of the virtualization vendor in November 2023. The vendor’s decision to remove a free tier, allegations of hard-sell tactics aimed at retaining subscribers, and several active court disputes over existing contracts — including ongoing perpetual licenses — have pushed many organizations to reassess core elements of their IT stacks.
Some rivals have benefited from that scrutiny. Nutanix, SUSE, and IBM have been named among firms seeing interest from customers reevaluating their dependence on VMware technology. At the same time, the realities of enterprise environments make wholesale exits difficult: many deployments are highly integrated and migration carries material financial and operational costs.
Large VMware environments tend to bind compute, storage, networking and orchestration pieces together. Pulling workloads out of those setups often requires retooling applications, retesting performance, adjusting storage and network topologies, and scheduling downtime windows. Those tasks add up in expense and risk to quality-of-service metrics, so a sizable number of customers opt to remain on known platforms rather than accept the uncertainty of a full migration.
Embedding AI at the core of a platform raises its own risks. Rearchitecting foundational software to place AI components in the critical path could produce breaking changes that affect live workloads. The deeper those architectural shifts go, the higher the chance that customer performance or reliability will be disrupted when updates land in production environments.
Broadcom’s initial pitch aims to soften that transition by making it easier to run models and agents inside existing VMware estates. It said VMware Private AI Services will arrive with VCF 9 subscriptions next year and will package components intended to help enterprises design, store and run AI on-premises or outside hyper-scale public clouds.
The announced bundle is expected to include a model store — with the expectation that many customers will experiment with smaller, open-source models during testing phases — plus indexing services and vector databases to support retrieval and similarity searches. An agentic AI builder is planned for composing autonomous workflows, while a prebuilt API gateway will handle optimized machine-to-machine communications for situations where distinct models must interact.
Speakers at the conference argued that AI usage within enterprises will expand and that platforms should reflect that trend. For now, Broadcom’s additions read as an incremental move into AI tooling rather than a radical reinvention. The company also outlined upgrades to the VMware Tanzu platform, such as simpler publishing of MCP servers and a new data lakehouse offering called Tanzu Data Intelligence aimed at modern data pipelines and analytics.
One of the more straightforward features on show was Intelligent Assist for VCF, a chatbot connected to VMware’s knowledge base. That assistant is designed to handle routine troubleshooting and triage, addressing queries and issues so human engineers are engaged less frequently in first-line support tasks.
There is precedent for predictions that a new pattern of infrastructure would sweep legacy approaches away. When containers surged in popularity some declared the end of traditional virtualization, and cloud proponents once forecasted the decline of on-premise databases and their large vendors. Reality has tended to be more mixed: existing investments in hardware, tooling and operational expertise often keep enterprises consolidated on platforms they already run, even when licensing costs are steep.
Broadcom appears to be adding AI-focused features to customer deals as a way to stay competitive and relevant amid that inertia. The company’s longer-term revenue model still depends heavily on the presence of legacy infrastructure at the hearts of enterprise datacenters, making steady subscription income more predictable than risky mass migrations.