Proxmox Backup Server vs Veeam: which one in 2026?
PBS vs Veeam in 2026: features, real 3-year TCO, measured performance, MSP scenarios. Which one fits your backup strategy?
Since Broadcom’s VMware acquisition and the climbing Veeam Universal renewals, the same question lands in our inbox every week from a CIO or an MSP: stay on Veeam, or switch to Proxmox Backup Server? There is no simple answer. After eighteen months migrating dozens of customers in both directions, we do have a grid that holds up.
This comparison rests on field numbers: 300+ PBS datastores monitored at Cloud-PBS, 80+ managed servers in production, Veeam B&R 12.x and 13 audits run for clients, and several documented two-way migrations. No marketing table copied from vendor docs. We look at the real 3-year cost, the deduplication ratios actually measured on real storage, the limits neither side advertises, and the scenarios where one solution clearly wins.
PBS and Veeam in two minutes
Proxmox Backup Server (PBS) is the native backup tool of the Proxmox ecosystem. Open source, GPLv3, built on the same Debian base as Proxmox VE. First stable release in 2020. PBS handles Proxmox VMs, LXC containers, and since 3.0 also Linux and Windows hosts through the proxmox-backup-client. The binary is free; a paid subscription unlocks the enterprise repository and contractual support. PBS 4.2, the current release in 2026, brings improved multi-datastore sync and native S3 object storage.
Veeam Backup & Replication (VBR) plays in a different category. It is a commercial suite designed for large hybrid infrastructures. It covers VMware vSphere, Hyper-V, Nutanix AHV, AWS, Azure, GCP, Microsoft 365, Salesforce, Kubernetes, and since the 12.3 branch Proxmox VE natively, extended in VBR 13 (released late 2025) to Proxmox VE 9 with VSS integration and application-level restore. Pricing follows, with a per-workload Universal License model. Veeam launched in 2008 and is now a pillar of the enterprise market.
From a distance, the two look like they do the same job. Up close, they reflect opposing philosophies: PBS is a minimalist backup engine purpose-built for Proxmox, Veeam is a multi-cloud data protection platform with fifteen years of features stacked on top.
Key features head to head
| Capability | PBS 4.2 | Veeam VBR 13 |
|---|---|---|
| Supported hypervisors | Proxmox VE, LXC | VMware, Hyper-V, Nutanix, Proxmox VE |
| Application backup (SQL, Exchange, AD) | Image-level only | Agent + application-aware |
| Item-level restore | Mount + manual extraction | Mailbox, AD object, SQL table in a few clicks |
| Object storage | Native S3 (PBS 4.x), Azure Blob via gateway | Native S3, Azure, GCS with immutability |
| Replication / sync | Sync between PBS datastores | CDP, Backup Copy, replica VM |
| Tape | Native LTO | LTO and VTL |
| SaaS protection (M365, Salesforce) | No | Yes (dedicated modules) |
| Cross-cloud restore | No | Direct Restore to AWS, Azure, GCP |
On pure virtualisation, PBS covers what most Proxmox-only shops actually need. If your fleet mixes vSphere, Hyper-V and Proxmox, or if you have to back up Microsoft 365 and Salesforce, Veeam stays ahead. That is its historic raison d’être.
The gap also widens on application-level recovery. Veeam Explorer for Microsoft Exchange restores an individual mailbox, even a single email, from any restore point. With PBS you restore the entire Exchange VM, boot it in isolated mode, and extract what you need. It works. It is slower. And it assumes you have the compute headroom available to spin up a 200 GB Exchange VM just to recover one PST.
On immutability, both deliver. Veeam relies on S3 object-lock or Hardened Linux Repositories. PBS encrypts chunks client-side, and ZFS datastores with immutable snapshots reach equivalent guarantees. To answer NIS2 or DORA on the “tamper-proof” axis, both pass, provided the storage layer is properly configured. The backup tool alone never delivers immutability; it is always tool + repository together.
The real 3-year cost
On paper, PBS is free and Veeam is paid. Reality is more nuanced because the cost of a backup system is never just the licence. For a typical fleet we work with, 30 Proxmox VMs, around a dozen TB of useful data, RTO 4 hours, RPO 12 hours, four cost lines actually matter over 3 years.
Hardware and target storage are equivalent across both solutions. Whether you put a PBS or a Veeam Backup Server in front, you buy the same class of server, the same disks, and the same external target for the 3-2-1 copy. This line does not discriminate the software choice.
The licence is what opens the gap. PBS is free in community mode, with optional Proxmox Server Solutions subscriptions per CPU socket for SLA-backed support. Veeam is billed per workload via the Universal License, with annual maintenance on top. On the 30-VM profile mentioned above, we observe a ratio of 1 to 6 or 1 to 8 between the two lines, in PBS’s favour. That is the main reason so many IT teams call us looking for an exit from a Veeam renewal.
The external S3 repository for the off-site copy is identical on both sides. If you turn on immutability on the Veeam side, plan for it on the PBS side too via ZFS snapshots or S3 object-lock. It is not a saving you can claim by switching to PBS.
Operating time is the argument Veeam reps lead with, and it is partially valid. Veeam saves admin hours on complex application restores (Exchange, SQL, AD) and on click-to-configure multi-site replication chains. PBS asks for slightly more CLI work on those scenarios. If your help-desk handles two or three application restores a week, that operating gap claws back part of the licence delta. If you restore four times a year, it weighs almost nothing.
Overall, on a 30-VM fleet, the 3-year TCO on the Veeam side runs about two to three times that of a properly sized self-hosted PBS. The gap shrinks on very small infrastructures (where Veeam licensing is lighter in absolute) and widens above 100 workloads.
Then there is the renewal question. The move to Veeam Data Platform Foundation has smoothed the increases, but on accounts inherited from the older Standard / Enterprise / Enterprise Plus editions, we regularly see significant jumps at the mandatory switch. When a customer calls us alarmed by a renewal quote, that mechanism is usually behind it, and that is what triggers the migration project, more often than a cold TCO calculation.
Measured performance and deduplication
Marketing decks show deduplication ratios between 10:1 and 30:1 on both sides. The field tells a different story.
PBS: ratios from 4.2:1 to 7.8:1 depending on the data profile. Across our 300+ monitored datastores (VictoriaMetrics, 20 s polling), the median measured at the end of 2025 sits at 5.4:1. The PBS engine relies on content-defined chunking (4 MB average chunks, SHA-256 hashing, zstd level 1 by default). Fast on ingest, less aggressive than the optimistic numbers some comparisons quote.
Veeam: ratios from 3:1 to 6:1 on standard backup with intra-job dedup. The ratio climbs to 8:1 or even 12:1 if you enable inline dedup on a ReFS or XFS repository with block cloning, and up to 20:1 on external Data Domain. But those high numbers assume a target that knows how to deduplicate. On a plain NTFS repository or a standard NFS share, you fall back to 3:1 or 4:1.
On restore speeds, here is what we measured on our own infrastructure:
- PBS on a ZFS RAIDZ2 datastore + NVMe SSD special device: 850 MB/s average live restore
- Veeam VBR 12.x on a well-tuned ReFS Backup Repository: 1.1 GB/s average (VBR 13 advertises roughly double on the same hardware; we do not have enough measurements yet to confirm)
- PBS via sync between remote datastores (10 Gb LAN): 600 MB/s
The Veeam-vs-PBS gap on restore comes mostly from ReFS tuning and the Veeam Proxy memory cache, not from any algorithmic superiority of the engine. Both solutions are bottlenecked by network and target storage long before the backup engine itself becomes the limit. If your RTO window is 4 hours on 12 TB, either solution holds, the difference plays out on the storage tuning layer, not on the software choice.
Ecosystem, integrations, MSP scenarios
The Veeam ecosystem is mature after fifteen years on the market. Veeam ONE plugins for monitoring, Veeam Backup for AWS/Azure/GCP as native agents at the hyperscalers, Service Provider Console with fine multi-tenancy and usage-based billing for MSPs, native integration in most SIEMs, official support at every major hosting provider. When you sell backup-as-a-service to 200 customers, Veeam’s MSP tooling has no real rival today.
The PBS ecosystem is younger but growing fast. The proxmox-backup-client has covered Linux and Windows hosts since 2023. Native datastore sync lets you build a hub-and-spoke topology without third-party tooling. On the MSP side, PBS multi-tenancy relies on namespaces (introduced in 2.2) and fine-grained per-user ACLs. Functional, technically clean, but less polished than the Veeam console when it comes to managing 200 tenants with automated billing.
For an MSP serving 50 SMB customers on Proxmox, PBS covers 80% of cases and costs five to ten times less in licences. For an MSP hosting multi-hypervisor or selling managed backup including SaaS protection, Veeam stays the most complete platform to monetise those value-added services. Among our 80+ Cloud-PBS managed servers, that first profile is what we see most: in-house IT teams or MSPs who have rationalised their infrastructure around Proxmox and want a native backup, without the Veeam invoice but without running PBS themselves either.
If you are not yet settled on the lower layer (vzdump vs PBS, the 3-2-1-1-0 rule, ZFS sizing), our complete Proxmox backup guide for 2026 covers everything upstream of this decision.
Limits to know before choosing
PBS has three limits you have to accept knowingly.
First, item-level Microsoft Exchange or SharePoint restore does not exist. You boot the entire VM in recovery mode and use Microsoft tools to extract what you need. On a 500-mailbox Exchange, that is 1 to 2 hours of recovery vs 3 minutes on Veeam Explorer. If your help-desk handles three “I deleted an email” tickets a week, it adds up.
Second, no SaaS protection. If you have to back up Microsoft 365, Salesforce or Google Workspace, you need a third-party tool (Afi.ai, AvePoint, Veeam Backup for M365 bought standalone, or sovereign options like Synology Active Backup for Microsoft 365). PBS does not cover that scope and probably will not in any visible roadmap.
Third, contractual support exists only through Proxmox Server Solutions Enterprise subscriptions, billed per CPU socket. The 2-hour P1 response is reserved to the top-tier plans. The Proxmox forum is excellent for peer help, but it is not an SLA.
Veeam has its shadows too.
The cost becomes absurd above 100 workloads. The Universal License grid climbs fast, discounts can be negotiated but stay far from free, and the annual maintenance bumps have become a political topic in many IT departments.
The VBK format is proprietary. Leaving Veeam after 5 years of backups is a project in itself: third-party conversion tools, sometimes a partial history rebuild. PBS uses a documented chunk-based format, descended from the Borg/Restic lineage, readable outside of PBS if needed. Portability is not a side argument in a long-term strategy.
The Veeam server footprint is heavier. A production Veeam Backup Server wants 8 to 16 cores and 16 to 32 GB of RAM minimum, plus dedicated Veeam Proxies depending on throughput. An equivalent PBS holds on 4 cores and 16 GB for the same volume.
If you are wrestling with the self-host vs managed-PBS question, we broke that arbitration down layer by layer in Managed PBS vs self-hosted PBS.
Verdict: which axis to arbitrate on
Choosing between PBS and Veeam is not a question of open-source-vs-proprietary ideology. It is a question of scope, volume, and application requirements.
Stay on Veeam if you run multi-hypervisor (vSphere + Hyper-V + Proxmox), if item-level Microsoft restore is central to your SLAs, or if your SaaS workload outweighs your VM footprint. It is the tool built for those scenarios and it does them better than any alternative today.
Switch to PBS if your fleet is Proxmox-first or about to become one, if your RTO/RPO is honest but not extreme, and if you want out of a licence cycle that doubles every three or four years. The benefit is not only financial, it is also the end of lock-in to a proprietary format.
A third path exists: hand the PBS operations over to a managed operator. That is what we do at Cloud-PBS for SMBs and MSPs who want the savings of a self-hosted PBS without running the hardware, the ZFS tuning, or the off-site copy themselves. If you are weighing a painful Veeam renewal against a full in-house rebuild, we offer a free audit of your current backup strategy: a one-hour conversation, a TCO comparison on your actual fleet, and an honest opinion on which path makes the most sense for you. The contact form is at cloud-pbs.com/contact.