Proxmox VE 9.2 adds dynamic load balancing and WireGuard SDN — VMware has run out of excuses for SMBs
Released on 21 May 2026, Proxmox VE 9.2 ships with a dynamic load balancer, WireGuard as an SDN fabric protocol, Ceph Tentacle, and kernel 7.0. Two years into the post-Broadcom exodus, the balance of power has definitively shifted.
In November 2023, Broadcom closed its $69 billion acquisition of VMware and killed perpetual licensing. In May 2026, Proxmox VE 9.2 shipped with a dynamic load balancer, WireGuard as an SDN fabric protocol, Ceph Tentacle, and kernel 7.0. By July 2026, the Proxmox fleet surpassed 1.5 million active hosts — and Broadcom had started losing Fortune 500 accounts, not just SMBs.
Version 9.2 does not just close the gap with VMware. It starts opening one of its own.
What VMware became under Broadcom
The licensing change was not a price hike — it was a different product forced onto existing customers. Before Broadcom, an SMB with 200 cores on vSphere Enterprise Plus paid roughly $40,000 to $80,000 per year. Under the new regime, that same cluster must move to VMware Cloud Foundation (VCF), the only available SKU in most cases. The bill: $200,000 to $400,000 per year. Five times the previous rate.
The contract changed in structure, not just amount. Perpetual licensing with optional support is dead. Everything is subscription — 1, 3, or 5 years — with a minimum of 16 cores per CPU socket, billed even if the machine runs at half capacity. Broadcom also slashed its partner program to a closed list and dropped entry-level SKUs like vSphere Essentials Plus with no equivalent replacement.
For SMBs, the message was clear by 2024. By 2026, it applies to large enterprises too.
Proxmox VE 9.2 — the technical leap
Released on 21 May 2026, version 9.2 runs on Debian 13.5 “Trixie”, ships kernel 7.0 as the stable default, and bundles QEMU 11.0, LXC 7.0, ZFS 2.4, plus both Ceph Squid 19.2.3 and Ceph Tentacle 20.2.1 in final form.
Six new capabilities redefine what an open-source hypervisor can do in production:
- Dynamic load balancer. The new balancer automatically migrates VMs between nodes to even out resource usage. It is not yet VMware DRS in full maturity — no advanced affinity rules — but the natural drift that plagues every cluster now corrects itself without manual intervention. For a 3-to-10-node cluster, this was the single most requested feature.
- WireGuard as an SDN fabric protocol. Proxmox SDN, introduced in 2023, now supports WireGuard as a native encrypted fabric. Connecting a multi-site cluster, a DR lab, or a hybrid cloud extension is as simple as a standard WireGuard tunnel — but managed from the Proxmox console. For organizations that want overlay networking without the complexity of NSX-T, this is a credible alternative.
- Advanced BGP and EVPN filtering. Route maps and prefix lists arrive in Proxmox SDN. This is not a homelab feature — it is a datacenter feature for environments serious about network segmentation.
- Custom CPU models in the GUI. Heterogeneous clusters — typical of test environments and migrations — can now create bespoke CPU models directly from the web interface. No more editing
/etc/pve/by hand to get Intel and AMD nodes to coexist in the same cluster. - HA arm/disarm. A single button in the Datacenter UI to freeze or ignore High Availability during maintenance windows. No one should have to fight their own cluster to shut down a node cleanly.
- Kernel 7.0 and Ceph Tentacle. The kernel jump brings broader hardware support and I/O optimizations. Ceph Tentacle 20.2.1 stabilizes distributed storage at scale — Proxmox is no longer just a hypervisor. It is a software-defined storage platform that holds its own against vSAN.
Why SMBs are switching — the real numbers
Talking about VMware-to-Proxmox migration without talking about money misses the point. Here are three documented scenarios:
| Scenario | VMware cost (post-Broadcom) | Proxmox cost | Annual savings |
|---|---|---|---|
| 3 hosts, ~40 VMs (SMB cluster) | ~$35,000/yr (VCF) | ~$1,500/yr (3 × Enterprise) | $33,500 |
| Lab/dev 12 hosts (CI/CD) | ~$80,000/yr (VCF) | $0 (Community) | $80,000 |
| 80 VMs production (mid-size) | ~$150,000/yr | ~$5,000/yr (Premium) | $145,000 |
The Community edition of Proxmox is production-ready at no cost, with no feature restrictions. The optional subscription starts at €110 per year per CPU socket (Community) and goes up to roughly €1,050 per socket (Premium), which includes technical support and priority access to the Enterprise repository.
A 6-week migration, engineering included, pays for itself in under 3 months.
What Proxmox still does not replace
Intellectual honesty demands naming the gaps.
The SDN is not NSX-T. WireGuard as a fabric protocol is a major step forward, but it does not replace micro-segmentation, distributed routing, or kernel-level virtual firewalls. If your VMware infrastructure depends heavily on NSX-T, those functions must move to a dedicated firewall (OPNsense, pfSense) or an external overlay solution.
The load balancer is not DRS. The 9.2 balancer handles post-placement rebalancing — it corrects drift but does not control initial placement with complex affinity rules. For 90% of clusters, this is enough. For the remaining 10%, it is a 2027 roadmap item.
The support ecosystem remains smaller. Proxmox Server Solutions GmbH has grown since 2024, but its commercial footprint is nothing like VMware’s in 2022 or Nutanix’s today. Large accounts requiring named support contracts and dedicated account managers should verify the Premium offering’s scope — it covers most cases, but every critical deployment warrants a review.
Proxmox Backup Server: the cluster’s insurance policy
A migration is not judged at cutover — it is judged at the first incident. Proxmox Backup Server (PBS) operates alongside Proxmox VE. Backups are incremental — only the delta between snapshots is transmitted — and deduplicated server-side. Encryption is applied client-side before transmission via AES-256-GCM.
In practice, a 3-node Proxmox cluster with a 4th node dedicated to PBS and remote replication achieves a resilience level comparable to vSphere + Veeam — at a fraction of the cost.
Verdict
If you have fewer than 500 cores, a baseline of in-house Linux skills, and no deep NSX-T dependency: migrate to Proxmox VE 9.2. The cost-to-capability ratio is so lopsided in Proxmox’s favor that staying on VMware is budget inertia, not a technical decision. With dynamic load balancing, WireGuard SDN, and Ceph Tentacle, 9.2 has removed the last three excuses that kept fence-sitters from moving.
If you have more than 500 cores or heavy NSX-T dependency, two options: pilot Proxmox on your dev/test environment for 12 months and migrate production at the next refresh cycle, or evaluate Nutanix AHV if your leadership requires an enterprise-grade commercial contract.
If you are 100% Windows with no Linux skills, Proxmox is still technically superior but your team is not ready. Train a Linux pair this year and postpone migration by 12 months — do not renew VMware for 3 years in the meantime.
The post-Broadcom migration wave is not a transient phenomenon. It is a market correction. Proxmox VE 9.2 is no longer the challenger — it is becoming the default, and this release proves why.
References
- Proxmox Virtual Environment 9.2 with Dynamic Load Balancer released, Proxmox Server Solutions, 21 May 2026.
- Proxmox VE 9.2 new features: Dynamic Load Balancer, SDN, and Ceph Tentacle, 4sysops, May 2026.
- Proxmox VE 9.2 Is Here: 7 New Features That Actually Matter, Virtualization Howto, 22 May 2026.
- Proxmox Virtual Environment 9.2 available!, Proxmox Support Forum, May 2026.
- Proxmox Backup Server Features, Proxmox Server Solutions.