Broadcom, the new owner of VMware, announced at VMware Explore that VMware Cloud Foundation (VCF) is now “AI native.” This matters because VMware underpins much of enterprise IT, and any shift in its architecture ripples out across industries. The timing is tricky. VMware has faced pushback since Broadcom acquired it in November 2023: the end of its free tier, changes in licensing, lawsuits around perpetual licenses, and reports of aggressive sales tactics. Many customers are unhappy, some are already looking at Nutanix, SUSE, or IBM.

The new AI positioning is Broadcom’s attempt to keep VMware relevant while customers calculate the cost of staying or leaving. VCF being AI native doesn’t mean VMware rebuilt everything. It means AI features will sit inside the existing environment so enterprises can run models and agents without jumping to hyperscale cloud providers.

What’s actually being offered

The most tangible piece is VMware Private AI Services, expected with VCF 9 subscriptions next year. It bundles the components needed to build and run AI workloads on-premise:

• A model store where organizations can pick up smaller, often open-source models, useful for testing phases.

• Indexing services and vector databases for handling embeddings and search.

• An agentic AI builder, pointing to multi-agent systems as a future enterprise use case.

• A ready-made API gateway designed for machine-to-machine communication between models.

There were also updates to the VMware Tanzu Platform: easier publishing of MCP servers and the launch of Tanzu Data Intelligence, a new data lakehouse. And one low-bar but practical feature: Intelligent Assist for VCF, a chatbot powered by VMware’s knowledge base, meant to delay how soon a ticket reaches a human support agent.

Why it matters

Enterprises don’t treat VMware as just another piece of software. It’s the backbone for heavily virtualised environments, often hosting mission-critical workloads. Migrating away isn’t easy. It’s costly, slow, and introduces risks to service quality metrics. Many enterprises would rather tolerate high licensing fees and controversial policy changes than risk destabilising workloads. This lock-in is VMware’s long-term safety net.

Adding AI features doesn’t remove that dependency. What it does is prevent VMware from looking dated. Enterprises experimenting with large language models and agents want some level of on-premise capability. If VMware couldn’t offer that, it would make migrations to alternative providers more attractive.

Risks and mistakes to avoid

The risk of “AI native” VMware isn’t theoretical. Re-architecting platforms at this depth can break workloads. Enterprises adopting the new features need to plan carefully. Test with small models first. Validate latency and integration impacts. Don’t assume the new services are production-ready out of the box. If AI features destabilise VMware stacks, the cost will be felt in downtime and in support escalations.

Another mistake would be treating VMware’s AI additions as unique. They’re not. The model store, vector databases, and agentic frameworks mirror what’s already available in open-source ecosystems and from cloud providers. The real advantage VMware offers is convenience inside existing VCF deployments.

The long-term view

Broadcom isn’t betting VMware’s future on AI. It doesn’t need to. Its business is secured by entrenched infrastructure that enterprises can’t easily unwind. AI is an add-on, a layer to keep VMware competitive against the noise of containers, hyperscale cloud, and alternative virtualization vendors.

For users, the takeaway is to evaluate VMware’s AI claims with caution. The long-term relationship with VMware is already defined by cost, licensing, and inertia. AI features may be useful in the short term, but they don’t change the structural reality: VMware knows its revenue is protected by legacy infrastructure, not by breakthrough innovation.