Mr.PlanB Logo
    Managed Cloud Kubernetes Ledger 2026
    Ledger
    Managed Cloud Kubernetes Ledger 2026

    Amazon Elastic Kubernetes Service: Cloud Kubernetes Ledger

    EKS is building toward hands-off clusters (Auto Mode, Capabilities) and very large AI training estates.

    Edition October 13, 20266 min read
    Open the interactive LedgerMethodology

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

    1. Snapshot

    • Tier: Hyperscaler
    • Owner: Amazon Web Services, part of Amazon.com, Inc. (listed, profitable)
    • Licence model: Pay as you go: $0.10 per cluster-hour, $0.60 in extended support; Auto Mode, Hybrid Nodes, Provisioned Control Plane and Capabilities billed extra
    • Products scored: EKS clusters with managed node groups, Fargate, Auto Mode, Hybrid Nodes, add-ons and EKS Capabilities
    • Latest release: Kubernetes 1.37 on 2026-10-01, 36 days after upstream
    • Next signal: EKS MCP Server in preview since 2025-11-21 with no GA date; open roadmap items carry no dates
    • Archetype: Broad, fast, paid by the hour
    • Evidence grade: A. Version calendar, document history, pricing, SLA, What's New posts and the containers-roadmap API were opened; compliance scope by year and partner counts were not re-verified.

    2. Scorecard

    DimensionScoreDirection
    Capability today86.0 / 100higher is better
    Momentum since end of 2022+18.2 pointshigher is better
    Credibility22.5 / 25higher is better
    Commercial risk8.0 / 25lower is better
    Technical lock-in10.0 / 25lower is better
    Integrator fit17.0 / 20higher is better
    AI leverage13.5 / 20higher is better
    Ledger Index80.7 / 100balanced weights

    3. Capability by domain

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

    Domain (weight)201620192022Today
    Versions, upgrades and support (12)n/a334
    Nodes, autoscaling and compute choice (12)n/a3.545
    Networking (10)n/a33.54
    Storage and data protection (8)n/a234.5
    Security and compliance (12)n/a3.544
    Observability and day-2 operations (8)n/a2.534.5
    Multi-cluster, hybrid and fleet (10)n/a1.534
    Developer and app platform (8)n/a1.534
    GPU and AI workloads (10)n/a234.5
    Regions, sovereignty and SLA (10)n/a344.5
    Total / 100not released52.667.886.0

    Year by year: 2018: 42.0 · 2019: 52.6 · 2020: 57.6 · 2021: 60.4 · 2022: 67.8 · 2023: 71.6 · 2024: 78.6 · 2025: 85.0 · 2026: 86.0

    What moved the score

    YearDomainChangePointsTrigger
    2019Multi-cluster, hybrid and fleet1 → 1.5+1.0Worker nodes on AWS Outposts 2019-12-03
    2019Networking2.5 → 3+1.0Private API endpoint 2019-03-19
    2019Nodes, autoscaling and compute choice2.5 → 3.5+2.4Managed node groups 2019-11-18, Fargate pods 2019-12-03, Windows nodes 2019-10-07
    2019Observability and day-2 operations1.5 → 2.5+1.6Control plane logs to CloudWatch 2019-04-04; Container Insights
    2019Regions, sovereignty and SLA2.5 → 3+1.099.9% SLA 2019-01-16 and region expansion through 2019
    2019Security and compliance2.5 → 3.5+2.4IAM roles for service accounts 2019-09-04 and private endpoint
    2019Versions, upgrades and support2.5 → 3+1.2Managed node groups 2019-11-18 with rolling node updates; three supported minors
    2020Developer and app platform1.5 → 2.5+1.6Managed add-ons from 2020-12
    2020Nodes, autoscaling and compute choice3.5 → 4+1.2Arm Graviton nodes GA 2020-08-17, Spot in managed node groups
    2020Regions, sovereignty and SLA3 → 3.5+1.0SLA raised to 99.95% on 2020-03-26
    2020Security and compliance3.5 → 4+1.2KMS envelope encryption of secrets 2020-03-05; Bottlerocket GA 2020-08-31
    2021Multi-cluster, hybrid and fleet1.5 → 2.5+2.0EKS Anywhere GA 2021-09-08 and EKS Connector
    2021Observability and day-2 operations2.5 → 3+0.8Managed Prometheus and Grafana; ADOT add-on 2022-04-21
    2022Developer and app platform2.5 → 3+0.8Third-party add-ons in AWS Marketplace 2022
    2022GPU and AI workloads2 → 3+2.0A100, Inferentia and Trainium nodes with EFA networking
    2022Multi-cluster, hybrid and fleet2.5 → 3+1.0Local clusters on Outposts 2022-09-19
    2022Networking3 → 3.5+1.0IPv6 pods 2022-01, security groups for pods, PrivateLink for the EKS API 2022-12-16
    2022Regions, sovereignty and SLA3.5 → 4+1.0Most commercial regions plus GovCloud and China
    2022Storage and data protection2 → 3+1.6EBS CSI driver as managed add-on 2022-03-31; EFS and FSx for Lustre drivers
    2023Networking3.5 → 4+1.0Network policy in the VPC CNI 2023-08-29; no managed mesh or multi-cluster networking
    2023Observability and day-2 operations3 → 3.5+0.8Cluster upgrade insights with deprecated-API checks 2023-12-20
    2023Storage and data protection3 → 3.5+0.8Mountpoint for S3 CSI driver GA 2023-11-27; CSI snapshot controller add-on 2023-11-17
    2023Versions, upgrades and support3 → 3.5+1.2Lag cut to 42 to 43 days from 1.27 (2023-05-24); extended support preview 2023-10-04
    2024GPU and AI workloads3 → 4+2.0Capacity Blocks for managed node groups 2024-07-01; AL2023 accelerated AMIs 2024-10-11
    2024Multi-cluster, hybrid and fleet3 → 3.5+1.0Hybrid Nodes 2024-12-01 join on-premises machines to cloud clusters
    2024Nodes, autoscaling and compute choice4 → 4.5+1.2Auto Mode 2024-12-01 with managed Karpenter; Karpenter 1.0 2024-08-14
    2024Observability and day-2 operations3.5 → 4.5+1.6Observability dashboard 2024-11-18; node monitoring and auto repair 2024-12-16
    2024Versions, upgrades and support3.5 → 4+1.2Extended support GA 2024-01-16, 26 months per minor; upgrade insights; median lag 41 days; no release channels or maintenance windows
    2025Developer and app platform3 → 4+1.6EKS Capabilities 2025-11-30: managed Argo CD, ACK and kro
    2025GPU and AI workloads4 → 4.5+1.0DRA 2025-07-11 and 100,000-node clusters 2025-07-16; GPU sharing only via the self-installed NVIDIA operator
    2025Multi-cluster, hybrid and fleet3.5 → 4+1.0EKS Dashboard across regions 2025 and managed Argo CD capability 2025-11-30; no multi-cluster services
    2025Nodes, autoscaling and compute choice4.5 → 5+1.2100,000-node clusters 2025-07-16; Capacity Blocks, warm pools 2026-04
    2025Storage and data protection3.5 → 4.5+1.6AWS Backup for EKS 2025-11-10 restores cluster, namespace or volume; FSx CSI add-on 2025-05-23
    2026Regions, sovereignty and SLA4 → 4.5+1.0European Sovereign Cloud with EKS 2026-01; 99.99% SLA tier 2026-03

    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
    2023-08Network policy in the VPC CNID4 NetworkingC
    2023-11EKS Pod IdentityD9 SecurityC
    2023-12Access entries replace aws-authD9 SecurityC
    2024-01Extended support priced at $0.60 per hourD12 Packaging and pricingP
    2024-09SageMaker HyperPod on EKSD13 AcceleratorsP
    2024-12Auto Mode GAD2 Compute and autoscalingV
    2024-12Hybrid Nodes GAD3 Fleet and hybridM
    2024-12Node monitoring and auto repairD8 Observability and operationsC
    2025-07100,000-node clustersD13 AcceleratorsM
    2025-07Dynamic Resource AllocationD13 AcceleratorsU
    2025-11AWS Backup for EKSD5 Storage and data protectionC
    2025-11EKS Capabilities: Argo CD, ACK, kroD6 Developer platformV
    2025-11Managed EKS MCP server previewD7 AI-assisted operationsM
    2025-11Provisioned Control PlaneD1 Versions and lifecycleC
    2026-01European Sovereign Cloud with EKSD10 Sovereignty and regionsM
    • Centre of gravity: Accelerators (3), Security (2), Networking (1)
    • Driver mix: Vision 13%, Customer need 40%, Market window 27%, Upstream 7%, Portfolio 13%

    5. Credibility: 22.5 / 25

    CriterionScoreBasis
    Cadence5Dated document history and a kept version calendar; 1.27 to 1.37 each reached EKS 36 to 44 days after upstream
    Roadmap transparency5Public containers-roadmap board with labels and issue history since 2018
    Say-do32 of 4 resolved promises GA within 12 months; restricted labels 'Coming Soon' since 2023-03, Fargate GPU never shipped for EKS
    Velocity5Several substantive launches a quarter: Auto Mode, Hybrid Nodes, Backup, Capabilities, Provisioned Control Plane
    Lifecycle stability4.526 months per minor with extended support, rollback within 7 days; forced upgrade at end of extended support

    Say-do record, 2023 to 2026: 6 promises with a forward signal 2023 to 2026: 2 GA within 12 months (extended support, Pod Identity agent), 2 missed (#1451 open 43 months, #88 closed without EKS delivery), 2 pending; 9 roadmap requests delivered after a median of about 20 months; most headline features ship on announcement day.

    6. Commercial risk: 8.0 / 25 (lower is better)

    CriterionScoreBasis
    Price and licence volatility3Extended support fee from 2024-04 at six times standard; Auto Mode, Hybrid Nodes and Provisioned Control Plane fees added 2024 to 2025
    Purchase constraints1Pay as you go with no minimum, but no free control-plane tier
    Channel and access0.5Self-service sign-up, public docs and pricing, worldwide regions
    Owner stability0.5Amazon, stable and profitable
    Cost of staying supported314 months standard; staying longer costs $0.60 per cluster-hour; control plane auto-upgraded after 26 months

    Security record of the product itself: none found: no cross-tenant or managed-control-plane flaw in EKS disclosed 2023 to 2026; 2026 bulletins on the network-policy agent (CVE-2026-86831) and EFS CSI driver (CVE-2026-85781) affect single customer clusters.

    7. Technical lock-in: 10.0 / 25 (lower is better)

    CriterionScoreBasis
    Formats1.5Conformant Kubernetes; Auto Mode NodeClass and Karpenter CRDs are provider-specific but workloads use standard APIs
    Export path2Free egress for leavers on request since 2024-03-05; Velero works; AWS Backup restores only into AWS
    Stack coupling3IAM access entries, Pod Identity, VPC CNI, AWS load balancers and EBS storage classes are the default path
    Hardware / cloud coupling2Hybrid Nodes and open-source EKS Anywhere and Karpenter run outside AWS, but the managed control plane is AWS only
    Skills and tooling coupling1.5kubectl, eksctl and Terraform; AWS IAM knowledge needed

    8. Integrator fit: 17 / 20

    CriterionScoreBasis
    Partner programme openness4AWS Partner Network with an EKS service delivery designation; partner count not re-verified
    Multi-tenancy and self-service4AWS Organizations and accounts, access entries per cluster, EKS Dashboard across accounts
    API and infrastructure-as-code5Official Terraform AWS provider, Cluster API provider, CDK constructs, eksctl and public APIs
    Skills and certification4Large practitioner pool and AWS certifications; no EKS-specific certificate

    9. AI leverage: 13.5 / 20

    StrandTodayLatest step
    AI workload support3.52025: DRA 2025-07-11, 100,000-node clusters, SOCI lazy image loading 2025-08-29; no managed inference add-on
    AI platform stack42024: SageMaker HyperPod with EKS orchestration GA 2024-09-10
    AI-assisted operations32025: Amazon Q troubleshooting in the EKS console GA 2025-11-21; advisory only
    Agent openness32025: Open-source EKS MCP server 2025-05-29; fully managed server in preview from 2025-11-21 with IAM and CloudTrail

    AI say-do: 1 preview pending (managed EKS MCP server, 10.7 months), 1 roadmap request delivered (DRA); the AI on EKS is HyperPod training orchestration plus advisory troubleshooting, not an acting agent.

    10. Reading

    EKS is building toward hands-off clusters (Auto Mode, Capabilities) and very large AI training estates. The record shows the fastest version cadence of the big three since 2023 and a steady flow of launches, but long-standing roadmap promises can sit for years. A buyer gets depth and scale at a price: no free tier, a sixfold fee in extended support and add-on fees that keep growing.

    11. Open questions and evidence caveats

    • Roadmap board status history could not be read (token lacks read:project); say-do counts rely on labels and close dates.
    • us-east-1 events (2020-11, 2021-12, 2023-06, 2025-10) disrupted EKS cluster operations; AWS published detailed summaries.
    • Compliance scope by year (FedRAMP, DoD levels) not re-verified; the earlier DoD IL6 claim is unconfirmed.
    • SageMaker HyperPod EKS date and the Amazon Q console feature rest on search snippets of What's New posts.
    • Auto Mode management fee is about 12% on top of EC2 in AWS's own example; check total cost against self-managed Karpenter.

    More from Managed Cloud Kubernetes Ledger 2026