What is driving everyone out of VMware?
Per-core subscription licensing after the Broadcom acquisition, which changed the arithmetic for a large number of organisations at once.
The result is a genuine migration wave, and the European options are unusually strong for it because most of them are open source with no per-core licensing at all. Proxmox VE is free with no feature limits and support subscriptions from about €115 per CPU per year — the entire cost structure is different, not a discount on the same model.
What matters most in choosing is the migration path rather than the feature comparison. Proxmox ships an ESXi import wizard, XCP-ng documents a V2V route from VMware, and OpenNebula supports VMware import. Without one of those, the project is a rebuild.
The second thing to check is what you actually used vSphere for. If it was VMs, clustering and shared storage, Proxmox covers it. If it was NSX-level networking, nothing here is a direct equivalent, and that gap needs designing around rather than discovering.
What does Proxmox include that VMware charges for?
High-availability clustering, distributed storage and backup — the three things that make a VMware estate expensive, all included with no feature limits.
Proxmox VE combines KVM for virtual machines and LXC for containers with HA clustering, Ceph distributed storage, built-in backup, software-defined networking and a REST API, and none of it is gated behind a tier. The free version is the full version.
The ESXi import wizard is what makes migration realistic rather than aspirational, and the community around it is large and active enough that most problems are already documented and answered.
Proxmox Server Solutions GmbH has operated from Vienna since 2005 under AGPLv3, with support subscriptions from about €115 per CPU per year for teams needing someone accountable. The honest gaps: the interface is less polished than vCenter, Ceph needs careful sizing and design to perform well, and there is no NSX-equivalent networking layer.
Which platform matches how your team already works?
Four different operating models sit in this category, and picking against your existing skills is the usual mistake.
A hypervisor with a management console, closest to the vSphere model: XCP-ng with Xen Orchestra as a separate management plane, live migration, continuous replication and strong built-in backup, with commercial support from Vates in Grenoble from about €950 per host per year.
A private cloud with tenants and APIs: OpenNebula for genuine multi-tenancy, multi-site and edge orchestration across KVM, LXC and Firecracker, Apache 2.0 from Madrid — and considerably lighter to operate than OpenStack.
Kubernetes everywhere: SUSE Virtualization runs VMs through KubeVirt with Longhorn storage and Rancher integration, so containers and VMs share one control plane. Or the full OpenStack API surface: Canonical OpenStack, with a managed option where Canonical operates the cloud for you.
When is OpenStack actually the right answer?
When you genuinely need its API surface, and rarely otherwise — which Canonical is unusually direct about.
OpenStack gives you a complete cloud API: tenants, quotas, networks, images, volumes, orchestration. If you are building something other teams consume programmatically, or repatriating from AWS and need comparable primitives, that surface is the requirement and nothing simpler substitutes.
If you want to run virtual machines on your own hardware, it is dramatically more complex than a hypervisor manager, and the operational cost is ongoing rather than one-off. Canonical's Sunbeam simplifies deployment considerably and a ten-year support commitment per release addresses the lifecycle question, with Ubuntu Pro from about $75 per node per year.
The managed option is the genuinely interesting part: Canonical operates the cloud for you, which converts an OpenStack team into a subscription. Canonical Ltd is UK-based, so adequacy rather than EU establishment — worth noting if procurement names EU member states.
What about storage and service providers?
Two specialists sit here for buyers the hypervisors do not serve.
StorPool is distributed NVMe block storage rather than a hypervisor, delivering consistent sub-millisecond latency with native drivers for Proxmox, OpenNebula, OpenStack and Kubernetes. It exists for the workloads where Ceph's latency profile is the problem — databases, VDI, anything where tail latency is what users feel. StorPool operates the storage for you, scales without disruption, and is priced per usable TB from Sofia. It needs NVMe and a well-specified network, and it is commercial where Ceph is free.
Virtuozzo is built for hosting companies and cloud service providers rather than internal IT: multi-tenancy and metering designed for reselling, high container density that improves margins directly, integrated distributed storage, a self-service portal and a DevOps PaaS layer. Virtuozzo International GmbH has nearly three decades of virtualisation experience from Schaffhausen under Swiss jurisdiction.
For internal enterprise IT, Virtuozzo is the wrong fit — its whole design assumes you are selling the capacity.