Proxmox VE: VM Platform Ledger 2026 Scorecard

Part of the VM Platform Ledger (edition 2026-09-30). How the scores work: methodology.

Snapshot

Type: product
Owner: Proxmox Server Solutions GmbH (private, Vienna)
Engine: QEMU/KVM and LXC on Debian
Licence: AGPLv3; optional support subscription per CPU socket per year, no feature gating
Latest release: 9.2 (2026-05-21); arm64 build 2026-08-05
Next milestone: Load balancer for non-HA guests; SAN snapshots out of preview (no dates published)
Archetype: Dependable builder
Evidence grade: A. Release notes, roadmap revision history and pricing snapshots were all read from primary sources.

Scorecard

Dimension	Score	Direction
Capability today	69.5 / 100	higher is better
Momentum since end of 2022	+19.3 points (2015 to 2022: +15.3)	higher is better
Credibility	18.0 / 25	higher is better
Commercial risk	6.5 / 25	lower is better
Technical lock-in	2.5 / 25	lower is better
Integrator fit	12.0 / 20	higher is better
AI leverage	5.5 / 20	higher is better
Ledger Index	68.7 / 100	balanced weights; see the methodology

Capability by domain, 2015 to today

Scores 0 to 5 against fixed anchors; total is weighted to 100.

Domain (weight)	2015	2019	2022	Today
Compute core (15)	2.5	3	3	4
Storage (12)	3	3.5	3.5	4
Networking / SDN (10)	2	2	2.5	3.5
Data protection / DR (12)	1.5	2.5	3	3.5
Migration-in (8)	0.5	1.5	1.5	3.5
GPU / AI (10)	1	2.5	2.5	3.5
Kubernetes (8)	0	0	0	1.5
Security / confidential (8)	2	2.5	3	3.5
Fleet / multi-site (9)	1	1.5	2	3.5
Automation / ecosystem (8)	3	3	3	3.5
Total / 100	34.9	46.3	50.2	69.5

Year by year: 2015: 34.9 · 2016: 34.9 · 2017: 40.1 · 2018: 43.1 · 2019: 46.3 · 2020: 48.5 · 2021: 49.3 · 2022: 50.2 · 2023: 52.8 · 2024: 56.8 · 2025: 68.0 · 2026: 69.5

What moved the score

Year	Domain	Change	Points	Trigger
2017	Data protection / DR	1.5 → 2.5	+2.4	5.0: asynchronous ZFS replication
2017	Migration-in	0.5 → 1.5	+1.6	5.0: command-line disk and OVF import
2017	Storage	3 → 3.5	+1.2	5.0/5.1: ZFS storage replication; Ceph Luminous
2018	GPU / AI	1 → 2.5	+3.0	5.3: passthrough and vGPU in the GUI
2019	Compute core	2.5 → 3	+1.5	5.4/6.0: 2-node quorum device, live migration with local disks
2019	Fleet / multi-site	1 → 1.5	+0.9	5.4: quorum device for 2-node clusters
2019	Security / confidential	2 → 2.5	+0.8	6.0: ZFS native encryption
2020	Data protection / DR	2.5 → 3	+1.2	Backup Server 1.0: incremental, deduplicated, encrypted
2020	Networking / SDN	2 → 2.5	+1.0	6.2: SDN (experimental)
2021	Security / confidential	2.5 → 3	+0.8	7.1: vTPM and Secure Boot for guests
2022	Fleet / multi-site	1.5 → 2	+0.9	7.3: offline mirror, cross-cluster migration by CLI
2023	Kubernetes	0 → 1	+1.6	Community CSI and Cluster API providers appear
2023	Networking / SDN	2.5 → 3	+1.0	8.1: SDN fully supported
2024	Automation / ecosystem	3 → 3.5	+0.8	8.3: webhooks; growing partner ecosystem
2024	Migration-in	1.5 → 3.5	+3.2	8.2: ESXi import wizard with live import
2025	Compute core	3 → 3.5	+1.5	9.0: HA affinity rules
2025	Data protection / DR	3 → 3.5	+1.2	8.4: backup provider API; Veeam, Commvault, Cohesity support
2025	Fleet / multi-site	2 → 3.5	+2.7	Datacenter Manager 1.0: multi-cluster management, cross-cluster live migration
2025	GPU / AI	2.5 → 3.5	+2.0	8.4: vGPU live migration; NVIDIA-supported hypervisor
2025	Kubernetes	1 → 1.5	+0.8	9.1: OCI images as containers
2025	Networking / SDN	3 → 3.5	+1.0	9.0: SDN fabrics (OpenFabric, OSPF); BGP in 9.2
2025	Security / confidential	3 → 3.5	+0.8	8.4/9.1: AMD SEV-SNP and Intel TDX (initial)
2025	Storage	3.5 → 4	+1.2	9.0: snapshots on shared SAN (preview); third-party storage plug-ins
2026	Compute core	3.5 → 4	+1.5	9.2: dynamic load balancer for HA guests; arm64 build

Direction, 2023 to 2026

Each release item carries one theme and one driver (V vision, C customer need, M market window, U upstream, P portfolio).

Date	Item	Theme	Driver
2023-03	7.4: rebalance on start, node maintenance mode	D1 Core compute	C
2023-06	8.0: Debian 12 base; Ceph enterprise repository included	D12 Packaging	U
2023-11	8.1: SDN fully supported; host Secure Boot	D3 Networking	C
2024-03	ESXi import wizard	D5 Migration-in	M
2024-04	8.2: backup fleecing, automated install	D4 Data protection	C
2024-08	Veeam supports Proxmox VE	D11 Ecosystem openness	M
2024-11	8.3: OVA import, webhooks, initial SEV	D5 Migration-in	M
2024-12	Datacenter Manager alpha	D10 Fleet/edge	C
2025-03	NVIDIA lists Proxmox as supported vGPU hypervisor	D6 AI/GPU infrastructure	C
2025-04	8.4: vGPU live migration, backup provider API	D11 Ecosystem openness	C
2025-08	9.0: Debian 13 base	D12 Packaging	U
2025-08	9.0: SAN snapshots (preview), HA affinity rules, SDN fabrics	D2 Storage	C
2025-11	9.1: OCI images as containers; Intel TDX	D8 Kubernetes	V
2025-12	Datacenter Manager 1.0	D10 Fleet/edge	C
2026-04	Backup Server 4.2: S3 datastores, encrypted sync	D4 Data protection	C
2026-05	9.2: dynamic load balancer; BGP and WireGuard fabrics	D1 Core compute	C
2026-05	Datacenter Manager 1.1: central install server	D10 Fleet/edge	C
2026-07	NVIDIA Mission Control collaboration	D6 AI/GPU infrastructure	V
2026-08	Official arm64 build for NVIDIA Grace and Vera	D1 Core compute	V
2026-09	24/7 support; North America subsidiary	D12 Packaging	M
2026-09	Omnissa Horizon Ready certification	D11 Ecosystem openness	M

Centre of gravity: Core compute (3), Packaging (3), Ecosystem openness (3)
Blind spots (no items): AI-assisted ops, Security
Driver mix: Vision 14%, Customer need 52%, Market window 24%, Upstream 10%, Portfolio 0%
Idea-to-need index: 0.16 (0 = demand-pulled, 1 = thesis-pushed)

Credibility: 18.0 / 25

Criterion	Score	Basis
Cadence	3.5	Two to three releases a year and a major every two years with Debian; dates not published in advance
Roadmap transparency	4	Public roadmap, git and bug tracker; but no dates, and the planned list went unedited for 18 months
Say-do	3	9 of 21 listed items shipped, 5 partial, 7 pending, none dropped; the DRS-style balancer took 43 months. Most headline features shipped without being announced
Velocity	4	Several substantive capabilities per release since 8.1
Lifecycle stability	3.5	About 3 years per major, in-place upgrades with a checker; no long-term-support option

Commercial risk: 6.5 / 25 (lower is better)

Criterion	Score	Basis
Price and licence volatility	1.5	Same four tiers since 2013; about +6% a year since 2023
Purchase constraints	0.5	Runs without a key; any quantity; all features in every tier
Channel and access	0.5	Open download; partner entry by application
Owner stability	2	Small founder-owned company; no outside owner, but size is a dependency
Cost of staying supported	2	Each major supported about 3 years; upgrades are free

Technical lock-in: 2.5 / 25 (lower is better)

Criterion	Score	Basis
Formats	0.5	Standard QEMU disk formats; plain-text VM config
Export path	1	No export wizard, but nothing to convert
Stack coupling	0	Any storage, standard Linux networking
Hardware / cloud coupling	0	Any x86 server; fully offline capable
Skills and tooling coupling	1	Standard Linux and KVM skills

Integrator fit: 12.0 / 20

Criterion	Score	Basis
Partner programme openness	4.5	Open programme; vendor also sells direct at list price; terms unpublished
Multi-tenancy and self-service	2	Pools, roles and realms exist; no quotas, self-service portal or metering
API and infrastructure-as-code	3	Everything is a REST call; Terraform and Ansible are community-maintained only
Skills and certification	2.5	Training with certificate of completion; no exam-based certification

AI leverage: 5.5 / 20

How the platform uses AI, scored per year like capability. Only GA counts.

Criterion	2019	2022	Today	Latest step
AI workload infrastructure	2	2	3.5	2025: NVIDIA-supported vGPU hypervisor; vGPU live migration (8.4)
AI platform stack	0	0	0	2015: None
AI-assisted operations	0	0	0	2015: None
Agent openness	2	2	2	2015: Full REST API with tokens; MCP servers are community-only

Year by year: 2015: 3 · 2016: 3 · 2017: 3 · 2018: 4 · 2019: 4 · 2020: 4 · 2021: 4 · 2022: 4 · 2023: 4 · 2024: 4 · 2025: 5.5 · 2026: 5.5

AI say-do, 2023 to 2026: 4 AI items announced; 3 GA, 0 still preview, 0 pending, 0 dropped. The fourth is an NVIDIA collaboration with nothing to check. Proxmox announces little AI and ships what it announces.

Proxmox runs AI workloads well enough since NVIDIA started supporting it in 2025, including vGPU live migration and Grace hosts. It has no AI platform, no AI in its own operations and no official agent interface; the community fills that gap with unofficial MCP servers.

Reading

Proxmox builds what its users have been asking for, in the order it can afford to, and the VMware exit pulled several of those items forward. It has closed most of the basic gaps since 2023 (SDN, import, load balancing, multi-cluster management, third-party backup) while staying the lowest-risk platform to buy and to leave. What it does not do is set direction: Kubernetes and AI-assisted operations are absent from both the product and the roadmap.

Open questions and items to re-score

SAN snapshots, DHCP address management and the new firewall are still technology previews; check before promising them.
The load balancer covers HA-managed guests only; non-HA balancing has been on the roadmap for 46 months.
Multi-tenancy for service providers needs third-party tooling (no quotas or self-service).
Vendor size: support was Vienna business hours only until the 24/7 service goes live on 2026-10-19.