Hosting providers leaving VCSP
Current customer contracts can still run. New commitment contracts and renewals cannot. We assess the estate and quote a move onto kcore before the term ends.
Per-socket licensing, hosts that boot the previous generation if an update fails, and Windows guests. Download the evaluation ISO, or hire us to move an estate from VMware or Proxmox.
The install walkthrough is further down this page. API and Terraform detail is in Docs.
Pick the one that matches the renewal in front of you.
Current customer contracts can still run. New commitment contracts and renewals cannot. We assess the estate and quote a move onto kcore before the term ends.
kcore is priced per CPU socket per year, not per core, and there is no 72-core order minimum. Compare that number with the renewal before you re-platform.
Each host update is one transaction. If it fails, the machine boots the previous generation. That is the reason to look at kcore for retail, plant, and remote racks.
Outcomes a renewal conversation can use. Stack detail stays in the docs.
Every host change is one transaction. If an update breaks the machine, it boots the previous generation. The host operating system is immutable.
Windows Server and Windows desktop guests run next to Linux. A mixed estate does not have to split platforms to leave VMware.
A physical GPU can be assigned to one VM. Licensed NVIDIA vGPU, where many VMs share one card, is not available yet.
VMs move between hosts when their disks are on shared Ceph storage. The control plane can replicate across datacenters. This is not vSphere HA, and live migration needs Ceph.
kcore-migrate inventories a vCenter or Proxmox estate and converts exported disks to qcow2 or raw. It does not pull disks straight from vCenter. Cutover of a whole estate is still a project.
You pay per CPU socket per year. The evaluation ISO is free. There is no per-core meter and no 72-core minimum.
Straight answers for people evaluating a VMware exit
Evaluation ISO and kctl on GitHub Releases. Windows guests. A physical GPU assigned to one VM. Atomic host updates and rollback. Multi-node clusters. Live migration when disks are on Ceph. Control-plane replication across datacenters. A provisioner configures servers over IPMI and Redfish and netboots kcore with UEFI HTTP boot.
kcore-migrate inventories VMware and Proxmox and converts exported disks. A direct vCenter disk read is not included. NVIDIA vGPU is not shipped. There is no integrated backup product yet. Immutable backups are planned. kctl backup copies the controller database.
This recording is the first-run flow: install and approve a node, add an operator SSH key, create a Debian 12 VM, find its IP, and SSH in. It is not the migration path.
Read the quickstart blog post or the operator guides in Docs.
A controller holds desired state; node agents run Cloud Hypervisor on each NixOS host; you operate the cluster with kctl and the same gRPC APIs.
Built on NixOS. The host filesystem is read-only. Updates swap the entire system atomically.
gRPC all the way down. The CLI and Terraform provider are just API clients.
Add nodes by booting the ISO and pointing them at the controller. They register automatically.
Local directories, LVM, or ZFS on the host. Ceph is the shared-disk option.
$ kctl get vms
NAME STATUS NODE MEMORY VCPU
production-app running node-01 4096MB 4
staging-db running node-02 8192MB 8
dev-web stopped node-01 2048MB 2
$ kctl describe vm production-app
Name: production-app
Status: running
Node: node-01
Memory: 4096 MB
vCPU: 4 cores
Storage: 50 GB (local-lvm)
Network: 192.168.122.100
Created: 2026-03-10 14:23:45
$ █Declarative APIs, Terraform, and an MCP server for AI agents. Deep reference material is in Docs—this homepage stays focused on operators leaving VMware.
Open documentationCommunity ISO free on GitHub Releases. Standard, Premium, and Enterprise per CPU socket for signed ISOs, the enterprise channel, and support.
Pricing and BuySend socket count, roughly how many Windows and Linux guests, and whether you are a hosting provider or an on-prem estate. We reply with whether kcore is a sensible target and what a migration project looks like.