Leaving VMware after Broadcom: OpenShift Virtualization, Nutanix or cloud, and what the move really takes

Broadcom now sells VMware only as per-core subscriptions, and the cheap ways around that are closing: Azure retires its pay-as-you-go VMware SKUs with the licence included at the end of this month, and vSphere 8 loses general support next October. If your renewal lands in the next year, price the exit before you sign, so the renewal buys you time instead of locking you in.

The first hit is the bill. Broadcom stopped selling perpetual licences and support renewals in December 2023 and moved everything to subscriptions. Its own knowledge base says you must license at least 16 cores for every CPU, whatever the chip actually has. AT&T said in court papers that it faced a proposed 1,050% rise. A German user group complained to the European Commission about increases of several hundred percent, and CISPE, a European cloud trade body, says some providers saw more than 1,000%.

The second hit is time. Some customers told The Register their quotes arrived a handful of days before the old contract expired, and a trade press report says late renewals carry a 20% surcharge. Leaving takes longer than most budgets assume. The same paper reported that Toshiba expects one to two years to move 2,200 VMs.

Gartner's analysts expect real movement, but slowly. One Gartner research VP said more than a third of VMware workloads will run somewhere else by 2028, and Gartner now expects 55% of enterprises to start proofs of concept to move away by 2029. A CloudBolt-sponsored survey reported by CIO.com in February 2026 found 86% of firms cutting back on VMware, yet only 2% had moved three quarters of their estate. Most people are in the middle, paying for VMware while they work their way off it.

For engineers: an expired support contract doesn't switch anything off. Broadcom's own KB says the entitlement stays, ESXi and vCenter keep working, and licensed features stay unlocked. You lose support cases and the right to new patches. In May 2025 heise reported that Broadcom sent cease-and-desist letters asking some perpetual-licence holders with lapsed support to remove updates installed after their contracts ended, so patching an unsupported estate quietly is a legal risk as well as a technical one.

The VDDKdefinitionVMware's disk library. Backup and migration tools use it to read VM disks fast, and Broadcom stopped offering it as a public download this summer. change gets the least attention and hurts migrations the most. Azure Migrate, Red Hat's Migration Toolkit for Virtualization, Nutanix Move and virt-v2v all use it to read VM disks quickly. Red Hat's docs say VMs on vSAN won't migrate without it. One practitioner in the middle of a large Azure Local migration told Hacker News they'd saved a copy before it disappeared, and guessed the agent-based alternative fails on 10% or more of VMs. Before you pick a tool, ask its vendor how they get VDDK today.

Anyone on AVS pay-as-you-go has until the end of this month. Microsoft is keeping AVS. It's the SKU with the licence baked in that goes. Staying means buying VCFdefinitionVMware Cloud Foundation, Broadcom's full VMware bundle. It's sold as a subscription per processor core, and you now need it to run VMware in the big clouds too. from Broadcom and putting the key in the portal, plus a separate Broadcom vDefend Firewall add-on if you use the NSX firewall beyond what an older reservation covers. Microsoft tells Broadcom about those registrations every month.

There's no single winner here. Most estates end up split: some VMs stay on VMware for a while, a group moves to a new on-premises platform, and the rest go to cloud or get retired. The table shows what each road actually involves.

OpenShift VirtualizationdefinitionPart of OpenShift, Red Hat's Kubernetes platform, that runs virtual machines next to containers on the same cluster. Under the hood it's the KubeVirt project. makes most sense when you already run containers or plan to. A VM becomes a Kubernetes object, so the same cluster, pipelines and policies cover both. Red Hat sells a VM-only edition, OpenShift Virtualization Engine, that allows unlimited VMs on each subscribed host but no application containers. Its migration toolkit can move a running vSphere VM warm, copying most data first and cutting over with a short stop, as long as Changed Block TrackingdefinitionA VMware feature that records which parts of a disk changed. Migration tools use it to copy a running VM first and move only the changes at cut-over. is on for every VM and every disk.

Nutanix is the closest match to how a vSphere team already works: a GUI-first platform with its own storage and the Move tool for migrations. Microsoft describes Move as a no-cost tool. Nutanix says Move can convert clusters that already run ESXi on Nutanix hardware to AHVdefinitionNutanix's own hypervisor, which runs the VMs on Nutanix servers. You'd use Nutanix Move to get VMs onto it. in place, installing VirtIO drivers and keeping IP and MAC addresses. If you've got a general Fibre Channel SAN, though, check the list first. Nutanix's external-storage support covers Dell PowerFlex and Pure FlashArray, and a generic SAN isn't on it.

Proxmox VE draws smaller teams because every feature is free and support costs from €120 to €1,100 per socket each year. Its import wizard connects to ESXi directly and doesn't need VDDK. One team of four posted that their move to six Proxmox clusters took eight months. Talking to stakeholders took longer than copying a 192 TB VM, and they said Fibre Channel multipathing and the Veeam integration were the rough edges.

Cloud is really two different moves. Running VMware inside a cloud moves the hardware problem but keeps the licence, since Azure, AWS and Google now all want your own VCF subscription. Moving to native cloud VMs ends the VMware bill. Someone still has to resize every VM and rebuild backup and operations around that cloud's own tools, and AWS offers Transform for VMware, which uses AI agents to map dependencies and plan the waves.

Cleveland Clinic took the second road in public. At Red Hat Summit it said it had renewed for three years to buy time, had moved 450 of its more than 10,000 VMs to OpenShift Virtualization, and aims for a 50% lower total cost before the next renewal. Those are vendor-stage numbers, so treat them as a direction rather than a promise.

Copying disks is the easy part. What makes VMware exits slow is everything plugged into vCenter, from backup jobs to the NSX firewall rules. Someone involved in Red Hat's virt-v2v work called a 12,000-VM customer medium to large but not unusual. Their estimate for a job that size was a scoping exercise, then perhaps one to three months of conversions, with each VM down from a reboot up to 12 hours depending on the strategy.

Backup is where many plans come unstuck. Veeam Backup & Replication 13 now protects Proxmox VE, Nutanix AHV and OpenShift Virtualization VMs, and it can restore vSphere backups into OpenShift or AHV. One practitioner pointed out that backup products still carry VDDK, so a cross-platform restore can work as a migration path if your normal tool can't read disks. Check yours against the target before the pilot, not after the first wave. Our OpenShift page covers the platform work, and our cloud migration service covers wave planning and cut-overs.

# Every VM with size and guest OS
Get-VM | Select Name, NumCpu, MemoryGB, ProvisionedSpaceGB, @{N='OS';E={$_.Guest.OSFullName}} | Export-Csv vms.csv
# Is Changed Block Tracking on? (warm migration needs it)
Get-VM | Select Name, @{N='CBT';E={$_.ExtensionData.Config.ChangeTrackingEnabled}}
# vSAN datastores (some tools can't read them without VDDK)
Get-Datastore | Where-Object {$_.Type -eq 'vsan'}
# Raw device mappings
Get-VM | Get-HardDisk -DiskType RawPhysical,RawVirtual | Select Parent, Name, DiskType

Teams get squeezed like this because the hypervisor was treated as permanent, so everything was built to depend on it. Whatever platform you land on, the same habits keep the next vendor change from becoming a crisis.

We help infrastructure teams through the whole thing, from reading the renewal quote and building the inventory to piloting the target platform and running the waves. We stay neutral on where VMs should land, because most estates end up using more than one platform. See our OpenShift engineers if you need people for the build, or start with a free assessment.