Seven VMware alternatives, scored for your scenario

Seven VMware alternatives, scored for your scenario
Diego Arias

Diego Arias

Senior DevOps Engineer

November 16 2026Solutions

Leaving VMware means choosing what runs the VMs next. This page scores seven alternatives (Proxmox VE, XCP-ng, OpenStack, Nutanix AHV, Hyper-V, OpenShift Virtualization and Harvester) in six scenarios, then gives each one’s licensing, support terms, backup options and migration path from VMware. The vendor pages behind every line are listed by platform in Sources at the end, all read on 16 November 2026. Nothing here was run, and no cell is a number of mine: each one is Wins, Works or Weak, with the line from the docs behind it.

A request runs from vRA through vCenter, NSX and an ESXi host into vSAN

The scorecard

One VM in the middle, and the hypervisors that could run it next

How a cell is scored:

  • Wins: what the scenario needs is in the product and its own licence, documented by the vendor.
  • Works: it can be done, with a limit, an extra paid product or an extra platform to run.
  • Weak: the docs show a gap for this scenario, or say nothing about it.

More than one alternative can win a scenario. Each scenario below gives the reason behind its Wins and Weak cells.

Small team, tight budgetData sovereigntyOne vendor for servers and storageA Windows-first teamEdge sitesAlready on Kubernetes
Proxmox VEWinsWinsWorksWorksWorksWeak
XCP-ngWinsWinsWorksWorksWorksWorks
OpenStackWeakWinsWeakWeakWeakWorks
Nutanix AHVWeakWorksWinsWorksWinsWorks
Hyper-VWorksWorksWorksWinsWorksWorks
OpenShift VirtualizationWeakWorksWorksWorksWorksWins
HarvesterWorksWorksWorksWorksWorksWins

The six scenarios

A small team on a tight budget

Proxmox VE and XCP-ng win: both are free to run with every feature, and both publish their support prices.

  • Proxmox VE, Wins. AGPLv3; the No-Subscription repository needs no key, though Proxmox does not recommend it for production. Per socket per year, net of VAT: Community EUR 120, Basic EUR 370, Standard EUR 550, Premium EUR 1,100.
  • XCP-ng, Wins. “All features are available for free”; the Essential plan covers up to 3 hosts for 2000 a year.
  • Nutanix AHV, Weak. No published price; the small option is NCI-Edge, licensed per VM, “restricted to small clusters running up to 25 concurrently powered-on VMs”.
  • OpenShift Virtualization, Weak. No published self-managed price: “Pricing varies based on your sizing and subscription choices”; the Virtualization Engine edition is sold per bare-metal node.
  • OpenStack, Weak. Free code, but “OpenStack control planes should also be deployed across at least three failure domains”.

Data sovereignty

XCP-ng, Proxmox VE and OpenStack win: open source, run where you choose, and each one’s vendor or foundation documents the case.

  • XCP-ng, Wins. Vates is in Grenoble, France, and “Vates VMS is based on a 100% open-source virtualization stack, without dependency on a non-European proprietary vendor”.
  • Proxmox VE, Wins. Proxmox is in Vienna, Austria, the code is AGPLv3, and Proxmox Offline Mirror serves hosts that “are completely air-gapped”.
  • OpenStack, Wins. Apache 2.0, and “OpenStack serves as the foundation for sovereign private and hybrid clouds”, with “scheduler-enforced data residency through availability zones and host aggregates”.

One vendor for servers and storage

Nutanix AHV wins: the hypervisor, the storage and the management come in one licence.

A Proxmox VE cluster whose node disks are pulled into Ceph

  • Nutanix AHV, Wins. AHV is “not sold or licensed as a standalone product” but built into NCI, and “Nutanix Prism is included with Nutanix Cloud Infrastructure at no additional cost”; external arrays work too, with Pure Storage FlashArray generally available since 2025-12-10. Limit: “Nutanix Flow is available under a separate license”.
  • OpenStack, Weak. Storage (Cinder) is one more project, and who supports the whole depends on which of the 22 listed distributions you pick.

A Windows-first team

Hyper-V wins: it is part of Windows Server, which the team already licenses.

  • Hyper-V, Wins. In both Standard and Datacenter editions of Windows Server 2025. Standard covers “2 virtual machines, plus one Hyper-V host per license”, Datacenter “unlimited virtual machines”; suggested MSRP for 16 cores: USD 1,176 Standard, USD 6,771 Datacenter. Limit: Storage Spaces Direct needs Datacenter.
  • OpenStack, Weak. No Microsoft tooling, and the official image guide only points to “the last available trial version of Windows Server 2012 R2”.

Whatever the hypervisor, Windows Server stays licensed per core, with “CALs … required for every user or device accessing a server”.

Edge sites

Nutanix AHV wins: small clusters at every site, managed from one place, under its own licence.

  • Nutanix AHV, Wins. “1-node, 2-node, and 3+-node options”, and “Nutanix Central makes it simple to monitor and manage tens or hundreds of edge locations all in one place”; NCI-Edge licenses small clusters per VM.
  • OpenStack, Weak. OpenInfra’s edge stack is StarlingX, which “is not a part of the OpenStack cloud infrastructure project”.

Already on Kubernetes

OpenShift Virtualization and Harvester win: both run the VMs on the Kubernetes cluster itself.

A VM settles into a pod on a Kubernetes cluster running KubeVirt

  • OpenShift Virtualization, Wins. It “has long been an included feature with all self-managed OpenShift editions”, “an operator included with any Red Hat OpenShift edition”. Limit: live migration needs shared storage with ReadWriteMany (RWX) access.
  • Harvester, Wins. “Built on Kubernetes”, with “Virtualization management with KubeVirt” and “Storage management with Longhorn”, under Apache 2.0. Limit: production nodes need 16 cores and 64 GB of RAM.
  • Proxmox VE, Weak. It runs KVM VMs and LXC containers; the admin guide describes no Kubernetes layer, so Kubernetes runs inside guests.

Getting the VMs out: import tools and VDDK

Broadcom completed its acquisition of VMware on 22 November 2023. It then stopped selling perpetual licences, so “offerings will solely be available as subscriptions or as term licenses”, with VMware Cloud Foundation and vSphere Foundation as the two offers.

Many import tools read VM disks through VMware’s Virtual Disk Development Kit (VDDK). Broadcom’s wording is that “access to the VDDK SDK is made available only to those who are part of the Technology Alliance Program (TAP)”; its public download pages stopped answering in August 2026, and Red Hat says “VMware VDDK images are currently unavailable to end-users via standard public downloads”. So check each tool before you plan the move:

The VMware products checking out as one Broadcom bundle

AlternativeOfficial import toolNeeds VDDK?
Proxmox VEImport wizard, ESXi 6.5 to 8.0 tested (guide)The guide does not mention VDDK
XCP-ngXen Orchestra V2V, ESXi 6.x to 8.x (guide)No: "nothing to download from VMware or Broadcom"
OpenStackvirt-v2v, Coriolis, MigrateKit (guide)The guide does not say
Nutanix AHVNutanix Move, free (DirectPath)Not for eligible configurations, since Move 6.3.3
Hyper-VWindows Admin Center VM Conversion, preview (guide); SCVMM 2025, offline onlyWindows Admin Center: yes, VDDK 8.0.3 only; the SCVMM page does not mention it
OpenShift VirtualizationMigration Toolkit for Virtualization (solution 7146995)Yes for vSAN; Red Hat cannot distribute the image
Harvestervm-import-controller add-on (VM import)The page does not mention VDDK

Keeping the self-service when the hypervisor changes

The hypervisor is half of what VMware did; vRA’s self-service on top is the other half. In Kestra, the request is a flow: its typed inputs are what the team fills in, one task creates and powers on the VM on Proxmox VE, and one runs the Ansible playbook that configures it. Both plugins are open source, Apache 2.0. If the hypervisor changes, only the create task changes; the inputs and the Ansible step stay.

One Kestra flow creates the VM on VMware, then Proxmox, then Hyper-V

No published blueprint does exactly this with open-source tasks only, so the flow below joins two of them, each task copied as published: provision_vm with the vm_name and vm_id inputs from Provision a VM and Register It in NetBox (its NetBox steps, which are Enterprise, left out), and ansible_configure with the vm_address and environment inputs from Provision VMware VMs with Ansible and ServiceNow (its vCenter clone, ServiceNow and Slack steps left out). Written for Kestra 1.3.39 or later.

id: proxmox-vm-self-service
namespace: company.team
inputs:
- id: vm_name
type: STRING
defaults: app-vm-01
- id: vm_id
type: INT
defaults: 201
description: Unused Proxmox VMID to assign to the new VM.
- id: vm_address
type: STRING
description: "Hostname or IP that Ansible will reach the new VM on."
- id: environment
type: SELECT
values: [ dev, staging, production ]
defaults: staging
tasks:
- id: provision_vm
type: io.kestra.plugin.proxmox.vm.Create
description: Create and power on the VM on Proxmox VE.
host: "{{ secret('PROXMOX_HOST') }}"
node: "{{ secret('PROXMOX_NODE') }}"
username: "{{ secret('PROXMOX_USERNAME') }}"
password: "{{ secret('PROXMOX_PASSWORD') }}"
vmId: "{{ inputs.vm_id }}"
vmName: "{{ inputs.vm_name }}"
cores: 2
memory: 4096
disk: "local-lvm:32"
powerOn: true
- id: ansible_configure
type: io.kestra.plugin.ansible.cli.AnsibleCLI
inputFiles:
inventory.ini: |
[app]
{{ inputs.vm_name }} ansible_host={{ inputs.vm_address }} ansible_user=ubuntu
configure.yml: "{{ read(secret('ANSIBLE_PLAYBOOK')) }}"
commands:
- ansible-playbook -i inventory.ini configure.yml --extra-vars "env={{
inputs.environment }}"
retry:
type: constant
interval: PT2M
maxAttempts: 3

Where it stops

  • The VM needs an operating system first. As published, the create task gives the VM a blank disk (the task’s defaults are “1 vCPU, 1024 MiB RAM, 8 GiB disk”), so Ansible has nothing to reach until the VM boots an OS, from the task’s osTemplate (“a cloud-init image or ISO on the node’s storage”) or a template. Neither blueprint shows that step.
  • Someone writes the flow first, in YAML, before anyone can fill in the inputs.
  • A custom form is Enterprise. Apps, which “let you wrap any Kestra flow in a custom UI”, are Enterprise Edition and Cloud only; in the open-source edition, the form is the flow’s inputs.
  • Dropping vRA may not cut the bill. Kestra’s own migration guide: “If vRA is part of a VCF bundle you will keep buying for the hypervisor, removing vRA may not reduce that invoice at all”.

The seven alternatives, side by side

Licence and priceSupportBackupWorth knowing
Proxmox VE 9.2AGPLv3; a subscription buys the Enterprise RepositoryEUR 120 to 1,100 per socket a year; 24/7 tickets on PremiumProxmox Backup Server 4.2; Veeam is a partnerThe load balancer is in technology preview; several clusters on one screen need Proxmox Datacenter Manager
XCP-ng 8.3 LTSFree, “all features are available for free”; Xen Orchestra mostly AGPLv3Vates VMS, from 2000 a year for 3 hosts to 1800 per host a year for 24/7Built into Xen Orchestra; Veeam since 13.1The load balancer and V2V are only in the XO 5 interface for now
OpenStack 2026.2Apache 2.0Through one of 22 commercial distributionsCinder volume backups, FreezerA minimal install is already five services
Nutanix AHV (NCI 7.6)Inside NCI, per CPU core; no public price24x7, P1 answered in 1 hour or 30 minutesNative DR down to 0 RPO; VeeamFlow networking is a separate licence
Hyper-V (Windows Server 2025)Per core with CALs: USD 1,176 Standard, USD 6,771 Datacenter for 16 coresPer incident, 2-hour first response for Severity A; supported until 2034Azure Backup Server; checkpoints are not a backupLoad balancing runs only when a node joins, unless you turn it on
OpenShift Virtualization 4.22In every self-managed OpenShift edition; no public priceStandard in business hours, or Premium 24x7OADP, whole VMs or single filesLive migration needs RWX shared storage
Harvester v1.9Apache 2.0; subscription with no per-CPU fees, no public priceStandard, or Priority with 24x7 for Severity 1 and 2NFS or S3, Longhorn volumes onlyUpgrading to v1.9.0 is not recommended yet; v1.8.x maintenance ends 2026-12-29

What the sources leave open

  • Prices behind a sales call. Nutanix, Red Hat and SUSE publish no self-managed price, Microsoft gives no price for its per-incident support, and Vates prints $ and € beside the same figures.
  • Gaps in the docs. No Proxmox page says whether its import wizard uses VDDK, and no SUSE page covers sovereignty or air-gapped installs.
  • Pages behind a login. Nutanix Move’s supported ESXi versions and the AHV load balancer are documented on Nutanix’s support portal, which was not read.

What this doesn’t cover

Hardware sizing, prices quoted by a sales team, performance, and any platform outside these seven. Every line above is what the vendor’s own pages said on 16 November 2026; check the linked page before you sign anything.

Sources

The main pages behind this post, all read on 16 November 2026.

VMware and Broadcom: Broadcom 8-K, Exhibit 99.1; VMware blog; ShapeBlue

Proxmox VE: Admin guide; Pricing; Migrate to Proxmox VE

XCP-ng and Xen Orchestra: Licenses; Vates pricing; V2V guide

OpenStack: Software; Distributions; VMware to OpenStack guide

Nutanix AHV: AHV; Software options; DirectPath

Hyper-V and Azure Local: Editions comparison; Pricing; Migrate

OpenShift Virtualization: Subscription guide; About, OpenShift 4.22; Solution 7146995

Harvester: Harvester overview; SUSE Virtualization; VM import

Kestra: io.kestra.plugin.proxmox.vm.Create; io.kestra.plugin.ansible.cli.AnsibleCLI; Migrating off vRA and vRO

Newsletter

Get Kestra updates to your inbox

Stay up to date with the latest features and changes to Kestra