In May 2025, reporting about Broadcom’s cease-and-desist correspondence to some former VMware perpetual-license customers raised an uncomfortable question: what can an organization continue doing after its support and subscription agreement expires? The answer requires a distinction that is easy to miss in alarming headlines. An old perpetual software license, the right to download later maintenance releases, the right to open support cases, and the right to upgrade to a new major product version are separate entitlements. Treating all four as a single “license” can lead to bad technical and purchasing decisions.
The reported letters focused on the use of updates and other support deliverables after the contractual support term, rather than describing a mechanism that automatically turns off licensed ESXi hosts when support ends. These are historical reports about particular correspondence, not a universal notice that every VMware customer has received. The wider question is still relevant in 2026: how should an infrastructure team establish what its existing version and contract permit, protect a supported security posture, and plan a migration or subscription transition without an operational outage?
What the 2025 reports actually established
Reports dated May 8, 2025 describe letters delivered to some VMware customers that had allowed support arrangements to lapse. The letters, as reported, asserted that continuing to use maintenance releases, patches, enhancements, and upgrades obtained after expiration could violate support and intellectual-property terms. Reporting also described a narrowly worded exception involving certain high-severity zero-day security patches. That exception should not be converted into a blanket statement that unsupported customers can download or install any security update at no cost. Read the actual agreement and the precise text of any customer-specific notice. The contemporaneous May 2025 licensing correspondence reporting is useful for the chronology, but it does not replace the signed customer contract.
Broadcom’s product transition was already under way when those reports appeared. Its earlier announcement discontinued new sales of perpetual VMware licenses and moved commercial offers toward subscriptions while saying that customers could continue to use their already purchased perpetual licenses. It also differentiated existing active support commitments from new renewals. The original VMware licensing transition announcement is important context because ending sales of a licensing model is not, by itself, retroactive cancellation of every older grant.
Separate perpetual software operation from update entitlement
Broadcom’s own technical guidance on perpetual VMware installations says expiration of Support and Subscription (SnS) does not revoke the underlying software entitlement for the licensed version. ESXi and vCenter Server can continue running, and existing VMs do not lose power simply because the support contract ends. The same guidance distinguishes those operational rights from access to the support portal for new patches, upgrades, and technical cases. Broadcom explains this distinction in its documentation on perpetual VMware operation after SnS expiration. Administrators should not interpret normal operation as proof of an ongoing right to every later download.
Check the exact product, release, edition, and purchase terms rather than relying on a description such as “we have Enterprise Plus.” Some entitlements are edition-specific, and later platform versions may be outside the perpetual grant. For example, February 2026 vSphere Enterprise Plus program terms describe limits associated with legacy perpetual-license upgrade rights. A perpetual installation of an earlier version does not automatically authorize an upgrade into a new major version or the use of unrelated licensed components.
Likewise, distinguish an expired perpetual support agreement from a subscription that has reached its own term or that is licensed by a different metric. The operating consequences and contractual rights are not interchangeable. An environment may also contain mixed populations of licenses, acquisitions, reseller contracts, lab hosts, and special entitlements. Consolidating these into one spreadsheet label without underlying documents creates hidden compliance risk.
Build a defensible entitlement and update inventory
Start by gathering purchase records, signed orders, support-contract dates, product and edition identifiers, license keys, and download history. Map them to the actual ESXi hosts, vCenter instances, add-ons, and management platforms in service. Record precise build numbers and the provenance of any patch or ISO used during the period under review. Someone who sees “VMware 8” in a dashboard still needs to know the patch level, download entitlement, upgrade history, and whether the relevant agreement covered the change at the time it occurred.
Use a read-only export wherever possible before planning remediation. The objective is to reconstruct a sequence, not quietly modify evidence. Keep host inventories, update logs, deployment records, vendor portal receipts, and change approvals in access-controlled storage. If a disputed change exists, do not rush to uninstall a package merely because an outside article says all patches must be removed. Removal can affect dependencies, security posture, data integrity, and availability. The organization’s procurement, licensing, security, operations, and legal specialists should evaluate the actual notice and applicable contract together.
For renewals and new subscriptions, compare total cost of ownership over a period that includes support, hardware refresh, operations, and migration. Per-core licensing assumptions, minimum commitments, bundle composition, and support scope vary by offering and contract. A public price estimate does not establish the charge owed by a particular enterprise. Request an itemized quote and a written inventory reconciliation when making the commercial decision.
Keep patch access and security engineering separate
An unsupported-but-still-running host is not automatically safe. Security controls must account for the possibility that the organization is not entitled to retrieve or install newly released fixes. Identify affected hypervisors, management appliances, and networking components, map published vulnerabilities to the specific builds in use, and prioritize exposure and exploitability. Isolation of management interfaces, privileged-access controls, network segmentation, monitoring, and reduced attack surface can lower some risks but do not reproduce a missing vendor patch.
Build a transparent exception record. For each system that remains on an unsupported build, identify its owner, business dependency, internet and management exposure, compensating controls, accepted risk, review date, and supported replacement path. A vague entry saying “waiting for VMware renewal” does not provide a security team with a mitigation plan. Conversely, downloading unlicensed updates from third-party mirrors is neither a reliable security strategy nor a compliant substitute for authorized software delivery.
Where support access has ended, seek official clarification from the vendor or the account team rather than inferring entitlement from whether a download URL still works. Availability of a file on a server is not proof of permission to use it. Document responses and tie any mitigation or migration to approved operational change control.
Evaluate migration as a workload and recovery problem
A decision to remain, subscribe, consolidate, or move to an alternative platform requires more than comparing hypervisor subscription prices. Inventory guest operating systems, CPU and memory reservations, virtual networks, storage protocols, snapshots, management agents, backup formats, replication mechanisms, monitoring dependencies, and licensing terms for software running inside the VMs. Resource performance, application supportability, and operational skills may be more important than the advertised cost of the hypervisor.
A migration rehearsal should include an independently restorable backup, test recovery and application validation, not only a successful virtual-disk conversion. Snapshots, clones, and templates are useful operational tools for rehearsals and provisioning, but snapshots are not independent backups and may depend on parent disk chains or platform-specific metadata. Test the recovery path on the destination platform, including identity, networking, authentication, database consistency, and time-sensitive integrations. Define a fallback that is executable within the permitted maintenance window.
For large estates, group applications by criticality and complexity. Pilot workloads with well-understood recovery procedures before business-critical clusters. Account for RDMs, raw-device mappings, shared-disk or clustering requirements, GPU pass-through, specialized drivers, and application licensing tied to host identifiers. The least-risk decision might be to maintain an existing installation briefly while support entitlement is resolved, rather than rush a high-dependency application onto untested infrastructure.
Take care with tool selection. A container platform is not simply a cheaper substitute for virtual machines: it changes operating-system isolation, state management, security responsibility, monitoring, and the application deployment model. Replatforming a database or tightly coupled Windows workload can be a development project, not just a storage copy. Compare technical scope independently from the licensing controversy.
How to communicate a licensing decision internally
An operationally useful decision memorandum should distinguish verified facts, vendor policy, customer-specific contract interpretation, assumptions, and unresolved questions. It should identify all material evidence and its date, the systems affected, the exposure if no action is taken, feasible alternatives, and the party accountable for signing off each risk. Avoid statements such as “all perpetual VMware installations are illegal” or “anyone with an old key can keep applying updates.” Both oversimplify a more specific set of contractual and operational rights.
For audit readiness, keep a change trail showing which personnel accessed vendor repositories, who approved downloads, how software was introduced into the environment, and which builds were used before and after maintenance. In multinational organizations, legal and procurement reviews may depend on geography, affiliate structure, and specific agreements. An engineer can supply precise technical evidence without making unsupported legal conclusions.
The enduring lesson from the 2025 cease-and-desist controversy is that software operation, support services, updates, and upgrade rights must be recorded separately. Once those distinctions are explicit, an organization can choose a subscription, a supported migration, or a controlled transition with a clearer understanding of cost, security exposure, technical dependencies, and compliance obligations. This article explains the evidence and decision process; it is not legal advice or a substitute for reviewing individual licensing terms.