Oracle OCI Kubernetes Engine: Cloud Kubernetes Ledger

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

1. Snapshot

Tier: Hyperscaler
Owner: Oracle Corporation, public company
Licence model: Pay as you go: basic clusters free without SLA, enhanced clusters $0.10 per cluster-hour with SLA, virtual nodes $0.015 per node-hour plus compute
Products scored: OKE basic and enhanced clusters with managed, self-managed and virtual nodes on OCI public regions
Latest release: Kubernetes 1.36.4 on 2026-10-01; 1.37.0 preview 2026-10-08
Next signal: Production 1.37.1 promised 'in a timely fashion'; no other dated roadmap
Archetype: Free-tier GPU cloud Kubernetes
Evidence grade: A. All 197 OKE release notes, the versions page with Wayback history, the price API and the Pillar SLA document read directly; compliance scope, sovereign-realm availability and incident history remain unverified.

2. Scorecard

Dimension	Score	Direction
Capability today	70.4 / 100	higher is better
Momentum since end of 2022	+15.8 points	higher is better
Credibility	16.5 / 25	higher is better
Commercial risk	6.0 / 25	lower is better
Technical lock-in	9.0 / 25	lower is better
Integrator fit	13.5 / 20	higher is better
AI leverage	7.5 / 20	higher is better
Ledger Index	66.3 / 100	balanced weights

3. Capability by domain

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

Domain (weight)	2016	2019	2022	Today
Versions, upgrades and support (12)	n/a	2.5	2.5	3
Nodes, autoscaling and compute choice (12)	n/a	2.5	3.5	4.5
Networking (10)	n/a	2.5	3.5	4
Storage and data protection (8)	n/a	1.5	2.5	4
Security and compliance (12)	n/a	2.5	3.5	4
Observability and day-2 operations (8)	n/a	1	2.5	3
Multi-cluster, hybrid and fleet (10)	n/a	1	1	1.5
Developer and app platform (8)	n/a	2.5	3	3.5
GPU and AI workloads (10)	n/a	0	2.5	4
Regions, sovereignty and SLA (10)	n/a	2	2.5	3.5
Total / 100	not released	37.0	54.6	70.4

Year by year: 2018: 34.8 · 2019: 37.0 · 2020: 40.4 · 2021: 51.6 · 2022: 54.6 · 2023: 62.6 · 2024: 63.6 · 2025: 67.2 · 2026: 70.4

What moved the score

Year	Domain	Change	Points	Trigger
2019	Networking	2 → 2.5	+1.0	Private worker nodes 2019-02, regional subnets and service gateways 2019
2019	Security and compliance	2 → 2.5	+1.2	Secrets encryption with Vault keys 2019-09; private worker nodes
2020	Observability and day-2 operations	1 → 2	+1.6	OCI Monitoring for clusters 2020-03 and API server audit 2020-07
2020	Regions, sovereignty and SLA	2 → 2.5	+1.0	OKE in US Government Cloud 2020-12-23
2020	Storage and data protection	1.5 → 2	+0.8	Block volume CSI plugin 2020-08
2021	Developer and app platform	2.5 → 3	+0.8	OCI Service Operator for Kubernetes 2021-08
2021	GPU and AI workloads	0 → 2.5	+5.0	GPU shapes for node pools 2021-05-20; no GPU sharing
2021	Networking	2.5 → 3	+1.0	VCN-native private API endpoint 2021-03, network security groups 2021-06
2021	Nodes, autoscaling and compute choice	2.5 → 3.5	+2.4	Cluster Autoscaler 2021-04, Arm and GPU shapes 2021-05, flexible shapes, capacity reservations 2022-02
2021	Observability and day-2 operations	2 → 2.5	+0.8	Node Doctor troubleshooting script 2021-07, work requests
2021	Security and compliance	2.5 → 3	+1.2	Private API endpoint 2021-03, image signing and verification 2021-04, RBAC mapped to IAM groups 2021-05
2022	Networking	3 → 3.5	+1.0	VCN-native pod networking CNI 2022-07, network load balancers 2022-03
2022	Security and compliance	3 → 3.5	+1.2	Customer-managed keys for boot and block volumes 2022-01
2022	Storage and data protection	2 → 2.5	+0.8	File Storage PVCs 2022-01, volume expansion and performance levels 2022; no snapshots
2023	Developer and app platform	3 → 3.5	+0.8	Managed add-on catalogue 2023-03, later Istio, ingress, Metrics Server, GPU operators
2023	Nodes, autoscaling and compute choice	3.5 → 4	+1.2	Virtual nodes and 2,000-node enhanced clusters 2023-03, preemptible capacity 2023-04, self-managed nodes 2023-06
2023	Observability and day-2 operations	2.5 → 3	+0.8	OKE service logs 2023-09, path analysis 2024-01; health observation 2026-03; no managed Prometheus or cost per namespace
2023	Regions, sovereignty and SLA	2.5 → 3.5	+2.0	Enhanced clusters 2023-03: 99.95% SLA in multi-AD regions; 50+ regions; EU, UK and Australian sovereign realms, OKE coverage not verified
2023	Security and compliance	3.5 → 4	+1.2	Workload identity 2023-03, credential rotation 2023-08, external OIDC 2024-10; no OKE bulletin series or runtime detection found
2023	Storage and data protection	2.5 → 3	+0.8	Volume snapshots 2023-06 and clones 2023-10
2023	Versions, upgrades and support	2.5 → 3	+1.2	On-demand node cycling 2023-05; enhanced clusters; lag down to 126 to 138 days; still no channels or extended support
2024	GPU and AI workloads	2.5 → 3	+1.0	More GPU shapes 2023 and 2024, NVIDIA device plugin add-on 2024-12
2025	Multi-cluster, hybrid and fleet	1 → 1.5	+1.0	Full Stack Disaster Recovery groups clusters across regions 2025-01; no multi-cluster manager found
2025	Networking	3.5 → 4	+1.0	Native ingress controller 2023, Istio add-on 2024-01, dual stack 2025-02, Zero Trust Packet Routing 2026-05; no Gateway API or eBPF option
2025	Storage and data protection	3 → 4	+1.6	OKE in Full Stack Disaster Recovery 2025-01, Ultra-High Performance volumes, raw block; Lustre PVCs 2026-03
2026	GPU and AI workloads	3 → 4	+2.0	NVIDIA GPU and Network Operator add-ons 2026-07, RDMA compute clusters 2026-06, AMD operators, GPU host repair 2026-10; DRA not found
2026	Nodes, autoscaling and compute choice	4 → 4.5	+1.2	Karpenter Provider for OCI 2026-04, 20,000 nodes per cluster 2026-07; Windows nodes not found

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).

Date	Item	Theme	Driver
2023-03	Enhanced clusters with financially backed SLA	D12 Packaging and pricing	C
2023-03	Virtual nodes and workload identity	D2 Compute and autoscaling	C
2023-06	Self-managed nodes and volume snapshots	D2 Compute and autoscaling	C
2024-08	Renamed OCI Kubernetes Engine	D11 Ecosystem openness	P
2024-11	Ubuntu worker nodes for GPU workloads	D13 Accelerators	M
2025-01	OKE in Full Stack Disaster Recovery	D5 Storage and data protection	P
2025-02	Dual-stack clusters and 5,000-node clusters	D4 Networking	C
2025-05	Preview releases for new Kubernetes minors	D1 Versions and lifecycle	U
2026-04	Karpenter Provider for OCI	D2 Compute and autoscaling	M
2026-07	20,000-node clusters and NVIDIA operator add-ons	D13 Accelerators	M
2026-08	Managed node rollback	D1 Versions and lifecycle	C

Centre of gravity: Compute and autoscaling (3), Accelerators (2), Versions and lifecycle (2)
Driver mix: Vision 0%, Customer need 45%, Market window 27%, Upstream 9%, Portfolio 18%

5. Credibility: 16.5 / 25

Criterion	Score	Basis
Cadence	4	Around 25 to 35 dated release notes a year and a versions page with planned and actual dates per patch
Roadmap transparency	2.5	Planned version dates and preview releases are published; no roadmap board, little direction beyond CloudWorld
Say-do	3	7 of 10 promises 2023-2026 shipped within 12 months (70%); node health checks took 18 months, two CloudWorld items not found
Velocity	4	Several features a quarter in 2025-2026: 5,000 then 20,000 nodes, Karpenter, dual stack, rollback, GPU operators
Lifecycle stability	3	About 12 to 14 months per minor, free in-place upgrades, no extended support; service renamed 2024

Say-do record, 2023 to 2026: 10 items 2023-2026 from planned dates on the versions page, x.0 preview releases and CloudWorld 2022 and 2024: 7 shipped within 12 months, 1 late (node health observation, 18 months), 2 not found in OKE notes (Cloud Guard container security, Logging Analytics integration). No public roadmap.

6. Commercial risk: 6.0 / 25 (lower is better)

Criterion	Score	Basis
Price and licence volatility	1	One change since 2023: paid enhanced tier added 2023-03 while basic clusters stayed free; no later fee changes found
Purchase constraints	0	Pay as you go; basic clusters free; Free Tier Ampere credits
Channel and access	0.5	Self-service sign-up, public docs and price API, 50+ regions
Owner stability	2	Oracle is profitable but carries heavy debt-funded AI data-centre spending (not re-verified from filings in this pass)
Cost of staying supported	2.5	Minors supported about 12 to 14 months with no extended support option; upgrades free

Security record of the product itself: None found: Oracle publishes no OKE security bulletin series; one OKE security release note (CVE-2020-10749 patch, 2020-07-30) and no cross-tenant or managed-control-plane flaw found 2023-2026.

7. Technical lock-in: 9.0 / 25 (lower is better)

Criterion	Score	Basis
Formats	1	CNCF-conformant 1.24 to 1.36; few OKE-specific CRDs beyond OCI Service Operator
Export path	1	Egress 10 TB per month free then $0.0085 per GB; Velero and Cluster API provider available
Stack coupling	2	Workload identity to OCI IAM, VCN-native CNI and OCI load balancer annotations
Hardware / cloud coupling	3.5	Runs only on OCI regions; no first-party OKE extension to other clouds or on-premises found
Skills and tooling coupling	1.5	kubectl and Terraform; OCI compartments and CLI are provider-specific

8. Integrator fit: 13.5 / 20

Criterion	Score	Basis
Partner programme openness	3	Oracle PartnerNetwork with OCI tracks; Kubernetes-specific partner facts not re-read
Multi-tenancy and self-service	3.5	Compartments and IAM policies per team; no OKE fleet or team-scope feature
API and infrastructure-as-code	4	oracle/oci Terraform provider 134 million downloads, Oracle-maintained Cluster API provider, SDKs and CLI
Skills and certification	3	OCI Architect and DevOps certifications; small OKE skills pool

9. AI leverage: 7.5 / 20

Strand	Today	Latest step
AI workload support	3	2026: GPU and Network Operator add-ons, RDMA compute clusters, GPU host repair 2026
AI platform stack	1.5	2025: OCI AI Blueprints installs inference blueprints into OKE; launch date and GA status not verified
AI-assisted operations	0.5	2018: No AI-assisted operations for OKE found
Agent openness	2.5	2025: Official open-source oracle/mcp OCI API MCP server 2025-07 runs OCI CLI including OKE; no OKE-specific or managed server

AI say-do 2023-2026: 2 tracked, 1 shipped within 12 months (Ubuntu images for GPU workloads), 1 late (node health checks). OKE's AI work is GPU infrastructure plumbing; no operator assistant and no OKE-specific MCP server.

10. Reading

Oracle is turning OKE into the control layer for very large GPU clusters on OCI, adding node counts, GPU operators and repair tooling at a fast pace since 2025. The record shows steady dated release notes but slow adoption of new Kubernetes minors (about 100 days), no extended support and no fleet or AI-assistant layer. Buyers get a free basic control plane and cheap egress, but a narrower managed feature set than the three largest clouds and coupling to OCI only.

11. Open questions and evidence caveats

Version lag median 104 days for 1.26 to 1.36, worst among the hyperscalers checked; 1.35 took 132 days.
No extended support: each minor ends about 12 to 14 months after release.
Basic clusters carry no SLA; only enhanced clusters at $0.10 per hour have one.
No OKE security bulletin series, so patch timing for runc, containerd and kernel flaws cannot be checked.
Incident history from press only (London heat 2022-07-19); OCI status history not readable.