Mr.PlanB Logo
    On-Prem Kubernetes Ledger 2026
    Ledger
    On-Prem Kubernetes Ledger 2026

    Giant Swarm managed platform: On-Prem Kubernetes Ledger

    Giant Swarm is building toward a Cluster API platform that it operates inside the customer's own provider account, with a Backstage portal and, since…

    Edition October 13, 20266 min read

    Part of the On-Prem Kubernetes Ledger (edition 2026-10-13). How the scores work: methodology.

    1. Snapshot

    • Tier: Niche
    • Owner: Giant Swarm GmbH, Cologne; independent, founder-led private company (funding history not confirmed by a primary source)
    • Licence model: Managed service sold by contract; platform repositories are Apache-2.0 on GitHub; no public price list
    • Products scored: Workload cluster releases for vSphere, Cloud Director and Proxmox VE (Cluster API), management cluster, Backstage portal, observability, security bundle, Agent Platform
    • Latest release: vsphere-36.1.0 on Kubernetes 1.36.5, 2026-10-06
    • Next signal: Kubernetes 1.37 release not published yet (upstream 2026-08-26); no dated roadmap
    • Archetype: Managed platform operator
    • Evidence grade: B. Release dates, components and roadmap items come from the vendor's release repository, docs and issue tracker; 2016 to 2022 capability rests on repository and component lists (C or B), and KubeVirt, backup, air-gap and pricing could not be confirmed. CNCF certification is current only for the AWS line (Kubernetes 1.35, 2026-08-31); the vSphere line was certified for 1.28 only (2024-03-12).

    2. Scorecard

    DimensionScoreDirection
    Capability today56.6 / 100higher is better
    Momentum since end of 2022+20.4 pointshigher is better
    Credibility17.5 / 25higher is better
    Commercial risk11.0 / 25lower is better
    Technical lock-in8.5 / 25lower is better
    Integrator fit12.0 / 20higher is better
    AI leverage12.5 / 20higher is better
    Ledger Index63.3 / 100balanced weights

    3. Capability by domain

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

    Domain (weight)201620192022Today
    Install, upgrade and lifecycle (12)1.52.52.53
    Bare metal, edge and on-prem infrastructure (10)1.5222.5
    Multi-cluster and fleet management (10)1.52.533.5
    Networking (10)12.52.53.5
    Storage and data protection (10)1112
    Security and compliance (12)122.53.5
    Developer platform (8)0.51.51.53.5
    Observability and day-2 operations (8)11.534
    VMs on Kubernetes (VMware exit) (10)0000.5
    GPU and accelerator workloads (10)0002.5
    Total / 10018.431.636.256.6

    Year by year: 2016: 18.4 · 2017: 18.4 · 2018: 18.4 · 2019: 31.6 · 2020: 31.6 · 2021: 31.6 · 2022: 36.2 · 2023: 37.8 · 2024: 47.8 · 2025: 52.8 · 2026: 56.6

    What moved the score

    YearDomainChangePointsTrigger
    2019Bare metal, edge and on-prem infrastructure1.5 → 2+1.0KVM, AWS and Azure releases (kvm v8.0.0, 2019-03-21)
    2019Developer platform0.5 → 1.5+1.6app-operator 1.0.0 and chart-operator app platform (aws v9.0.0, 2019-10-28)
    2019Install, upgrade and lifecycle1.5 → 2.5+2.4kvm v9.0.0 and aws v9.0.0 (2019-10-28): versioned release bundles, operator-driven upgrades; no support window found
    2019Multi-cluster and fleet management1.5 → 2.5+2.0cluster-operator and app-operator manage clusters and apps from the management cluster (aws v9.0.0, 2019-10-28)
    2019Networking1 → 2.5+3.0Calico 3.9.1 in aws v9.0.0 (2019-10-28), flannel-operator on KVM
    2019Observability and day-2 operations1 → 1.5+0.8vendor on-call operation; no exporters in kvm v8.0.0 component list
    2019Security and compliance1 → 2+2.4cert-operator issues cluster certificates (kvm v8.0.0, 2019-03-21); no policy engine or scanner found
    2022Multi-cluster and fleet management2.5 → 3+1.0Flux GitOps for clusters and apps (vendor post 2021-11-09)
    2022Observability and day-2 operations1.5 → 3+2.4cert-exporter, net-exporter, etcd resource-count exporter and vertical pod autoscaler in kvm v16.2.1 and aws v18.0.0 (2022)
    2022Security and compliance2 → 2.5+1.2Kyverno app since 2021-03-19; PodSecurityPolicy removal post 2022-08-25
    2023Developer platform1.5 → 2.5+1.6Backstage plugin first release v0.1.0 (2023-07-19)
    2024Bare metal, edge and on-prem infrastructure2 → 2.5+1.0vSphere (vsphere-27.0.0, 2024-10) and Cloud Director (2024-11-07) on Cluster API; Proxmox app 2025-02-21; no bare-metal provider and no air-gap procedure in the docs
    2024Developer platform2.5 → 3+0.8Backstage developer portal with cluster and deployment views; 'Developer Platform MVP' roadmap epic closed 2024-10-02
    2024Install, upgrade and lifecycle2.5 → 3+1.2vsphere-27.0.0 (2024-10-16) joins the Cluster API release framework; two majors maintained; median lag to upstream 290 days, no published support period
    2024Multi-cluster and fleet management3 → 3.5+1.0Cluster API management clusters, organizations for tenants, kubectl-gs GitOps templates; hundreds of AWS clusters live-migrated (post 2026-04-01)
    2024Networking2.5 → 3.5+2.0Cilium with kube-proxy replacement and Hubble in vsphere-27.0.0 (2024-10-16); Envoy Gateway app 0.1.0 (2024-12-19); no multus or SR-IOV in the docs
    2024Observability and day-2 operations3 → 3.5+0.8Observability Platform on Mimir, Loki and Tempo introduced 2024-09-11
    2024Security and compliance2.5 → 3+1.2security bundle with Trivy Operator CIS reports, Kyverno Pod Security Standards, Falco, cosign policies shipped with Cluster API releases
    2024Storage and data protection1 → 2+2.0provider CSI default storage classes (vSphere CNS, Cloud Director disks); no bundled storage, no backup product in the docs
    2025GPU and accelerator workloads0 → 2.5+5.0GPU Operator app (2025-03-11, v1.0.0 2025-10-06), GPU node pools from release 30.1.0 (2025-03-18), AI Conformance 1.33 (2025-11-05); AWS and Azure only, not vSphere or Proxmox
    2026Developer platform3 → 3.5+0.8portal plus Agent Platform section; Flux HelmRelease replaces the App resource (docs 2026-07-03); no built-in CI or serverless
    2026Observability and day-2 operations3.5 → 4+0.8multi-tenant Grafana, OTLP ingestion, alert rules and silences; mean time to acknowledge about two minutes (docs)
    2026Security and compliance3 → 3.5+1.2structured OIDC authentication option in vsphere-34.0.0 (2026-01-22), Policy API; no FIPS or STIG in the docs
    2026VMs on Kubernetes (VMware exit)0 → 0.5+1.0KubeVirt on bare metal offered as a service in blog posts of 2026-04-21 and 2026-05-28; no docs, release or roadmap item

    4. Direction, 2023 to 2026

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

    DateItemThemeDriver
    2023-07Backstage plugin for the Giant Swarm portal, first releaseD6 Developer platformM
    2024-03Platform 3.0 on Cluster API; VMware Cloud Director addedD1 Install and lifecycleU
    2024-09Observability Platform on Mimir, Loki and TempoD8 Observability and operationsV
    2024-10vSphere joins the release framework (vsphere-27.0.0)D2 Bare metal, edge and infrastructureM
    2024-12Envoy Gateway app for Gateway APID4 NetworkingU
    2025-02Proxmox VE cluster app first releaseD2 Bare metal, edge and infrastructureM
    2025-03GPU node pools and NVIDIA GPU Operator app (release 30.1.0)D13 AcceleratorsC
    2025-06AWS vintage to Cluster API migration completedD1 Install and lifecycleP
    2025-11Kubernetes AI Conformance certification (1.33)D13 AcceleratorsM
    2026-01Optional structured OIDC authentication (vsphere-34.0.0)D9 SecurityU
    2026-04KubeVirt on bare metal offered as a service (blog)D10 VMs on KubernetesM
    2026-06Agent Platform v1.0.0D7 AI-assisted operationsV
    2026-07mcp-kubernetes and muster 1.0.0D11 Ecosystem opennessV
    2026-09Customer-facing AI chat in the portalD7 AI-assisted operationsC
    • Centre of gravity: Install and lifecycle (2), Bare metal, edge and infrastructure (2), Accelerators (2)
    • Driver mix: Vision 21%, Customer need 14%, Market window 36%, Upstream 21%, Portfolio 7%

    5. Credibility: 17.5 / 25

    CriterionScoreBasis
    Cadence3.5app and patch releases weekly; Kubernetes minors reach vSphere a median 290 days after upstream (1.27 to 1.36); numbering rule published but no calendar
    Roadmap transparency4.5public roadmap tracker (4,199 issues), public release notes per release, Apache-2.0 platform repositories; no dated roadmap
    Say-do39 of 15 roadmap items delivered within 12 months of opening (60%); vendor states no dates
    Velocity42024 to 2026 added vSphere, Cloud Director, Proxmox, EKS, Gateway API, GPU, Agent Platform; vsphere-36.1.0 on 1.36.5 in October 2026
    Lifecycle stability2.5KVM line ended 2022, vintage AWS to Cluster API migration finished June 2025, App resource deprecated 2026; no support period published

    Say-do record, 2023 to 2026: From the public roadmap (research record section F): 15 items counted, 9 within 12 months, 6 later or overdue (GPU nodes 726 days, registry stage 1 544 days, private clusters with proxy 503 days, SLO alerting 405 days, CAPA in China 406 days, SELinux by default open 406 days), 2 too young to count, 0 dropped.

    6. Commercial risk: 11.0 / 25 (lower is better)

    CriterionScoreBasis
    Price and licence volatility1no price or licence change found since 2023; prices not published
    Purchase constraints3managed service by contract only; no software-only purchase or self-serve trial
    Channel and access2.5docs and source public, partner page lists a few partners, no tiers
    Owner stability2independent founder-led company; 11% staff reduction on 2023-03-30; funding history unconfirmed
    Cost of staying supported2.5upgrades included; a release is deprecated once two newer majors exist; Flatcar and Cilium fixed

    Security record of the product itself: No CISA KEV entry, no NVD record and no published repository advisory for Giant Swarm components found on 2026-10-13 (KEV file dated 2026-10-08; 1,202 repositories searched); no adjustment.

    7. Technical lock-in: 8.5 / 25 (lower is better)

    CriterionScoreBasis
    Formats1.5conformant Kubernetes, Cluster API and Flux; Giant Swarm Organization and Release resources; App resource being retired
    Export path1.5customer owns the provider account and clusters; no documented exit procedure
    Stack coupling2Flatcar images and Cilium on every cluster; provider CSI
    Hardware / cloud coupling2vSphere, Cloud Director, Proxmox, AWS, Azure only; vendor engineers need provider API and outbound internet access
    Skills and tooling coupling1.5kubectl, Flux and Cluster API skills plus Giant Swarm release and portal tooling

    8. Integrator fit: 12.0 / 20

    CriterionScoreBasis
    Partner programme openness2.5partner page with a few integrators; no tiers or public terms
    Multi-tenancy and self-service3.5organizations per team in one management cluster; multi-tenant observability; Kamaji control planes added in cluster-vsphere 2026
    API and infrastructure-as-code3.5Cluster API, Flux HelmRelease, Helm charts and kubectl-gs; Terraform provider stale since 2020
    Skills and certification2.5training catalogue in the docs; no public certification

    9. AI leverage: 12.5 / 20

    StrandTodayLatest step
    AI workload support32026: AI Conformance 1.34 (2026-05-21), KubeRay and Kueue tutorials, model-manager in the Agent Platform
    AI platform stack22026: Agent Platform v1.0.0 (2026-06-05): agent runtime and model management; no training or MLOps platform
    AI-assisted operations32026: portal AI chat answering cluster questions through live tool calls, docs 2026-08-31, roadmap item closed 2026-09-08
    Agent openness4.52026: mcp-kubernetes v1.0.0 (2026-07-02) and muster v1.0.0 (2026-07-14): OAuth 2.1 single sign-on, user identity carried to the Kubernetes API, agent calls through one gateway

    AI say-do: 4 announced, 3 GA, 0 preview, 1 pending, 0 dropped. The AI is GPU support, an MCP gateway with user-level identity, an agent runtime and a read-only portal chat; no AI training platform.

    10. Reading

    Giant Swarm is building toward a Cluster API platform that it operates inside the customer's own provider account, with a Backstage portal and, since 2026, an agent platform on top. The record shows steady delivery and an open roadmap, but Kubernetes minors reach vSphere a median 290 days after upstream and the on-premises product has no documented bare-metal, air-gap, backup, VM or GPU support. A buyer gets an operated, open-source stack on VMware or Proxmox with low technical lock-in, at the price of a bespoke contract, a slow upgrade cadence and a small partner ecosystem.

    11. Open questions and evidence caveats

    • Kubernetes 1.37 has no release 48 days after upstream; the vSphere product was CNCF certified only for 1.28 (2024-03-12), the AWS product for 1.35 (2026-08-31).
    • KubeVirt, backup, FIPS, air-gapped operation and bare metal appear only as claims in blog posts or not at all in the docs; the earlier research record's versions of these were invented and removed.
    • No price list or support period in months is published; funding history is unconfirmed.
    • Say-do is measured from roadmap issue opening dates because the vendor states no delivery dates.

    More from On-Prem Kubernetes Ledger 2026