A Full VMware Exit Costs More Than Three Years of Broadcom’s Bill

Broadcom closed its purchase of VMware on November 22, 2023. Within weeks the company killed new perpetual licenses and folded a long catalog of products into per-core subscriptions. VMware Cloud Foundation became the anchor bundle, with VMware vSphere Foundation below it. Renewal quotes landing in front of infrastructure teams today look nothing like the ones from 2022.

So the question is simple. Does it cost less to leave? We ran the math a buyer would run, over three years, across three paths.

Staying Put Wins the First Two Years

Renewing is the cheapest path in year one, and it is not close. The subscription price rises, sometimes sharply. But the hardware, the runbooks, and the staff all stay where they are. No migration tooling. No downtime window. No retraining budget.

The bill is still real. Buyers report per-core subscription prices several times their old support rates, and Broadcom now pushes three-year commitments. Pin down your own number before you compare anything. A renewal that doubles still beats a rebuild that spends year one on design alone.

Bundling is the quiet part. VMware Cloud Foundation rolls vSphere, vSAN, NSX, and Aria into one line item. If you only ever wanted vSphere, you still pay for the rest. Cheaper tiers exist, and buyers say the gap between them narrows at every renewal.

Support is part of the value and part of the price. You keep one vendor, one queue, and a team that already knows the product. That continuity has a dollar value, and spreadsheets often drop it.

Hardware refresh is the catch. If your vSphere estate sits on servers due for replacement, you buy the new hardware anyway. You just buy it on Broadcom terms.

The Middle Path Looks Cheap and Rarely Is

A partial migration moves some workloads to a KubeVirt based platform and leaves the rest alone. Teams call this the pragmatic answer. It is the messiest one.

You now run two control planes, two backup tools, two patch cycles, and two sets of skills. The subscription drops for the workloads that moved. The overhead climbs everywhere else. A CloudBolt survey of 302 IT decision-makers found most are reducing VMware slowly, workload by workload. That fits. Slow reduction is expensive reduction.

Tooling is where the hidden math bites. You need a converter, a validation harness, and a rollback plan for every wave. None of that is free, and none of it retires when the migration ends.

Migration tooling and staff retraining show up here first. Training a virtualization team on Kubernetes storage and networking takes months, and the learning lands during production incidents, not in a lab.

A Full Exit Only Pays Off If You Count Year Four

Full exits go to Proxmox, SUSE Virtualization, Red Hat OpenShift Virtualization, or a hyperscaler. Each one trades license cost for labor and risk.

Year one is the worst. You pay for tooling and consultants while the old estate still runs. Year two, you retire the first wave and the run rate drops. Payback usually arrives in year three or later, and it stays thin until the last cluster is gone.

Hardware refresh lands twice on this path. You buy for the new platform while the old servers run out their final year. Support changes shape too. One throat to choke becomes a contract per platform, each with its own escalation path and its own gaps.

Downtime risk is the number nobody budgets. A failed cutover costs revenue, and the biggest workloads always move last, when the team is tired and the deadline is close. Measure the cost of one bad hour before you start. That number, not the license line, decides whether the exit clears its bar.

Staying on VMware is cheaper for three years in most estates. Leaving is cheaper only if you plan past that, absorb the retraining, and accept the risk. Analysts put large-scale migrations at twelve to eighteen months of work. Pick your path with that on the table.

The economics behind this migration wave, including the three year cost model and why exits run long, are in the 2026 State of Enterprise Infrastructure report.

Every announced commitment in this space is tracked with its source in our AI data centre power commitments record.

Related reading. The VMware Exit Has a Shortlist. Nutanix Just Took a Front Seat.. Broadcom Sold You Two VMware Bundles. You Probably Need One.. Bundling Is Not Unifying. Multicluster Kubernetes Management Still Has a Gap.. The 2026 State of Enterprise Infrastructure.