Tencent Kubernetes Engine: Cloud Kubernetes Ledger

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

1. Snapshot

Tier: Asian hyperscaler
Owner: Tencent Holdings (listed in Hong Kong), cloud unit Tencent Cloud
Licence model: Cluster management fee by cluster model since 2022-03-21 (L5 USD 0.0204 to L5000 USD 4.40 per hour); nodes and super-node pods billed as CVM or pod resources
Products scored: TKE general (managed) clusters with native nodes, super nodes and registered nodes, on Kubernetes 1.30 to 1.34
Latest release: Kubernetes 1.34 GA 2025-11-25; 1.36 CNCF-certified 2026-08-28
Next signal: Kubernetes 1.36 GA, policy says about three months after upstream (upstream 2026-04-22)
Archetype: Conservative Asian cloud Kubernetes
Evidence grade: A. Most scores rest on dated English update notes (2016 to mid-2025), the version schedule, billing and SLA pages read this run; later notes, compliance scope and AI tooling are thin.

2. Scorecard

Dimension	Score	Direction
Capability today	70.0 / 100	higher is better
Momentum since end of 2022	+3.8 points	higher is better
Credibility	13.5 / 25	higher is better
Commercial risk	9.0 / 25	lower is better
Technical lock-in	9.5 / 25	lower is better
Integrator fit	12.5 / 20	higher is better
AI leverage	7.0 / 20	higher is better
Ledger Index	58.6 / 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)	1.5	3	3.5	3.5
Nodes, autoscaling and compute choice (12)	1.5	3	4	4
Networking (10)	1.5	2.5	3.5	3.5
Storage and data protection (8)	1	2.5	3	3.5
Security and compliance (12)	1	2.5	3	3.5
Observability and day-2 operations (8)	1	2.5	3.5	4
Multi-cluster, hybrid and fleet (10)	0.5	1	2.5	3
Developer and app platform (8)	1.5	3	3.5	3.5
GPU and AI workloads (10)	0	2	3.5	3.5
Regions, sovereignty and SLA (10)	1	2.5	3	3
Total / 100	21.2	49.2	66.2	70.0

Year by year: 2016: 21.2 · 2017: 21.2 · 2018: 21.2 · 2019: 49.2 · 2020: 52.0 · 2021: 58.8 · 2022: 66.2 · 2023: 68.0 · 2024: 69.2 · 2025: 70.0 · 2026: 70.0

What moved the score

Year	Domain	Change	Points	Trigger
2019	Developer and app platform	1.5 → 3	+2.4	Image registry triggers, TencentHub CI, Helm applications 2018
2019	GPU and AI workloads	0 → 2	+4.0	GPU clusters 2018-05 and gpu_manager virtualization 2019
2019	Multi-cluster, hybrid and fleet	0.5 → 1	+1.0	CCN network registration of clusters
2019	Networking	1.5 → 2.5	+2.0	GlobalRouter, VPC-CNI beta 2019-06, CCN registration, private master access
2019	Nodes, autoscaling and compute choice	1.5 → 3	+3.6	Cluster autoscaler and HPA 2018-03, GPU scale-out, bare-metal clusters, 5,000-node quota 2019-10
2019	Observability and day-2 operations	1 → 2.5	+2.4	Log collection to CLS, Kafka or Elasticsearch, GPU metrics 2019-10
2019	Regions, sovereignty and SLA	1 → 2.5	+3.0	International console 2017, Singapore and other overseas regions; 99.95% SLA from 2019-07-04
2019	Security and compliance	1 → 2.5	+3.6	CAM, tag-based authorization and managed-cluster ACLs
2019	Storage and data protection	1 → 2.5	+2.4	CBS CSI open-sourced 2018, CFS and COS volumes 2019-12
2019	Versions, upgrades and support	1.5 → 3	+3.6	Managed masters 2019-04, even minors only, in-place rolling node updates 2019-12
2020	Networking	2.5 → 3.5	+2.0	VPC-CNI GA 2020-03, CLB direct to pod, IPVS-bpf beta, NetworkPolicy add-on 2020-12; no Gateway API or IPv6 confirmed
2020	Storage and data protection	2.5 → 3	+0.8	CBS-CSI with snapshots and online expansion 2020-12
2021	Developer and app platform	3 → 3.5	+0.8	App market, add-on management 2020-08, OLM add-on 2021-05; Crossplane in marketplace 2024
2021	GPU and AI workloads	2 → 3.5	+3.0	MIG 2021-07 and qGPU sharing with isolation 2021-09
2021	Multi-cluster, hybrid and fleet	1 → 2.5	+3.0	Registered clusters 2021-10 and external nodes 2021-06
2022	Nodes, autoscaling and compute choice	3 → 4	+2.4	Node pools, super nodes for serverless pods, external nodes 2021-06; ARM only in beta 2021-03
2022	Observability and day-2 operations	2.5 → 3.5	+1.6	Cloud-native Prometheus monitoring, event and audit dashboards, node self-healing, Request Recommendation 2021-10
2022	Regions, sovereignty and SLA	2.5 → 3	+1.0	SLA for all paid cluster models from 2022-03-21; eleven regions outside mainland China incl. Frankfurt, Riyadh, Sao Paulo
2022	Security and compliance	2.5 → 3	+1.2	KMS plug-in 2020-05, fine-grained RBAC, audit dashboard, image signature verification 2022-09
2022	Versions, upgrades and support	3 → 3.5	+1.2	In-place major upgrades 2020-08; 27-month support per even minor
2023	Multi-cluster, hybrid and fleet	2.5 → 3	+1.0	Registered IDC nodes over public network 2023-11; no fleet management or multi-cluster services found
2023	Storage and data protection	3 → 3.5	+0.8	Backup Center 2023-04, cross-cluster restore 2024-04, CBS encryption 2025-03
2024	Security and compliance	3 → 3.5	+1.2	KMS Secret encryption in managed clusters 2023-06, policy management 2024-01; workload identity and compliance scope not confirmed
2025	Observability and day-2 operations	3.5 → 4	+0.8	TKE Insight cost per namespace 2023-05, deprecated-API pre-checks, control-plane monitoring 2025-02

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
2023-04	Backup Center for backup, restore and migration	D5 Storage and data protection	C
2023-05	TKE Insight cost per workload and namespace	D8 Observability and operations	C
2023-06	KMS Secret encryption in managed clusters	D9 Security	C
2023-11	qGPU withdrawn from normal nodes	D13 Accelerators	P
2024-01	Policy management against cascade deletion	D9 Security	C
2024-04	Kubernetes 1.28 GA, 257 days after upstream	D1 Versions and lifecycle	U
2025-04	AIBrix multi-node inference guide	D13 Accelerators	M
2025-08	TKE Serverless cluster creation stopped for super nodes	D2 Compute and autoscaling	P
2025-11	Planned upgrades with maintenance windows; Kubernetes 1.34 GA	D1 Versions and lifecycle	C
2026-07	Cluster Security Enhancement for AI-agent workloads	D9 Security	M

Centre of gravity: Security (3), Accelerators (2), Versions and lifecycle (2)
Driver mix: Vision 0%, Customer need 50%, Market window 20%, Upstream 10%, Portfolio 20%

5. Credibility: 13.5 / 25

Criterion	Score	Basis
Cadence	3.5	Dated monthly English update notes 2016 to mid-2025 and a release schedule table; English notes stop after June 2025
Roadmap transparency	2.5	No public roadmap; English documentation lags and is partly machine-translated
Say-do	2	Version policy promises GA about three months after upstream; met for 1 of 5 minors (1.28 at 257 days)
Velocity	3	Steady incremental node, storage and policy features; few large launches since 2023
Lifecycle stability	2.5	27 months per even minor, but serverless clusters, qGPU on normal nodes, TencentHub, TPS and NginxIngress updates withdrawn

Say-do record, 2023 to 2026: 6 version-policy items 2023-2026: 1 on time (1.34 at 90 days), 4 slipped (1.26 at 138, 1.28 at 257, 1.30 at 125, 1.32 at 196 days), 1 pending (1.36 certified 2026-08-28). No other dated promises found; first-pass ledger of 9 shipped items with invented dates removed.

6. Commercial risk: 9.0 / 25 (lower is better)

Criterion	Score	Basis
Price and licence volatility	3	Free managed clusters became paid by cluster model on 2022-03-21, quotas imposed 2022-04-30, serverless clusters withdrawn 2025
Purchase constraints	1.5	Pay as you go from USD 0.02 per hour; no free managed tier
Channel and access	1.5	Self-service international site with English docs, some machine-translated; script access blocked on the main doc domain
Owner stability	1.5	Tencent is profitable and listed; cross-border regulatory exposure for non-Chinese buyers
Cost of staying supported	1.5	27-month support included; forced upgrades possible after expiry; even minors only

Security record of the product itself: No TKE-specific cross-tenant or managed-control-plane flaw found for 2023-2026 (NVD, TKE vulnerability pages, web search); runc CVE-2024-21626 and kernel CVE-2026-31431 fix pages require customer node action. No deduction.

7. Technical lock-in: 9.5 / 25 (lower is better)

Criterion	Score	Basis
Formats	1.5	CNCF-certified for each released even minor; workloads use standard APIs
Export path	2	Backup Center with cross-cluster restore since 2024-04; egress at Tencent rates
Stack coupling	2.5	VPC-CNI, CAM, CLB and CBS integration; TKE re-creates cloud resources declaratively
Hardware / cloud coupling	2	Registered clusters and registered IDC nodes; no first-party on-premises distribution counted here
Skills and tooling coupling	1.5	Standard kubectl and Terraform; tccli for cloud parts

8. Integrator fit: 12.5 / 20

Criterion	Score	Basis
Partner programme openness	2.5	Partner programme not checked this run; limited presence outside Asia
Multi-tenancy and self-service	3	CAM, projects and tags; registered clusters; no fleet layer
API and infrastructure-as-code	4.5	Partner Terraform provider tencentcloudstack/tencentcloud with 46M downloads, Cloud API 3.0, tccli
Skills and certification	2.5	Tencent Cloud certifications exist; not checked this run

9. AI leverage: 7.0 / 20

Strand	Today	Latest step
AI workload support	2.5	2025: AIBrix distributed inference and LLM deployment guides 2025-04, open-source components
AI platform stack	1	2021: Cloud-native AI service 2021-10; TI platform integration not confirmed
AI-assisted operations	1.5	2021: Rule-based health checks and Request Recommendation 2021-10; no AI assistant found
Agent openness	2	2019: Cloud API 3.0 with all-region access and tccli; no official MCP server found

AI say-do: 0 announced, 0 GA, 0 preview, 0 pending, 0 dropped; TKE's AI offer is GPU sharing plus documented open-source inference stacks, with no operator assistant or MCP server found in English docs.

10. Reading

Tencent runs TKE as a stable, even-minor Kubernetes service with long 27-month support and incremental node, storage and policy features, consolidating serverless into super nodes. The record shows versions arriving well after its own three-month target and no public roadmap or AI operations tooling, while managed clusters became paid in 2022. Buyers in Asia get a mature, inexpensive service with long support, weighed against slow version adoption, thin English documentation and limited reach elsewhere.

11. Open questions and evidence caveats

English update notes stop at June 2025; 2025-2026 capability may be understated.
GA dates come from the TKE schedule page; grayscale availability is proxied by CNCF certification dates.
No official MCP server or AI operations assistant found; Chinese-only pages were not searched exhaustively.
The 2024-04-08 control-plane outage (about 87 minutes) is known from press summaries of Tencent's postmortem.
Compliance programmes covering TKE were not verified.