Hybrid Kubernetes has moved from buzzword to buying decision. Organizations running workloads across cloud providers and on-premises data centers need a platform that handles the complexity without creating new problems. But the options vary wildly in what they actually manage, what they cost, and how much operational burden they leave on your team.
This post compares five platforms that compete for the hybrid Kubernetes market in 2026: Microsoft Azure Arc, Google GKE Enterprise (formerly Anthos), Amazon EKS Anywhere and EKS Hybrid Nodes, Red Hat OpenShift, and Cloudfleet. Each takes a fundamentally different approach, and the right choice depends on where you are starting from and what you need.
What to look for in a hybrid Kubernetes platform
Before comparing specific products, it helps to establish the criteria that matter most for hybrid deployments:
- Managed vs. self-managed: Does the platform handle the full cluster lifecycle (control plane, upgrades, node provisioning), or does it add a management layer on top of clusters you operate yourself?
- Multi-cloud support: Can you run workloads across multiple cloud providers and on-premises infrastructure from a single control plane, or do you need separate clusters per environment?
- Operational complexity: How much Kubernetes and cloud-specific expertise does your team need? How many agents, operators, or additional tools are required?
- Pricing transparency: Can you predict costs before you deploy, or do hidden add-ons, mandatory support tiers, and per-vCPU surcharges make the real price hard to estimate?
- Data sovereignty: Where does the control plane run? What metadata leaves your environment? Is the vendor subject to extraterritorial data access laws like the US CLOUD Act?
With these criteria in mind, let’s look at each platform.
1. Microsoft Azure Arc
Azure Arc is a management overlay that connects existing Kubernetes clusters to the Azure portal. It provides visibility, policy enforcement via Azure Policy, GitOps configuration through Flux v2, and integration with Azure services like Entra ID and Microsoft Defender.
What it does well. Arc is useful for organizations that already have Kubernetes clusters and want centralized governance through the Azure portal. It supports any CNCF-certified Kubernetes distribution, so it can connect clusters running on AWS, GCP, or on-premises infrastructure. The base pricing of $2 per vCPU per month (with 6 free vCPUs per subscription) is straightforward for the management layer itself.
Where it falls short. Azure Arc does not create clusters, provision nodes, manage Kubernetes upgrades, or handle networking. You are responsible for the entire infrastructure lifecycle. Arc installs 10-11 agents per cluster, requiring a minimum of 850 MB of memory and roughly 7% of a CPU. Air-gapped environments are not supported. If connectivity to Azure is lost for more than 90 days, the managed identity certificate expires and you must delete and recreate the Arc resource from scratch.
The real cost goes beyond the $2/vCPU base price. Most production deployments need Microsoft Defender for Containers (~$6.87/vCore/month), Azure Monitor Container Insights (usage-based, often $200-1,000+/month), and potentially Microsoft Sentinel for SIEM. For an 80-vCPU cluster, the combined management cost can reach $900 to $1,700 per month before any infrastructure expenses.
Best for: Organizations with existing Kubernetes clusters that want centralized Azure governance without changing their infrastructure.
2. Amazon EKS Anywhere and EKS Hybrid Nodes
Amazon offers two products for hybrid Kubernetes: EKS Anywhere, which runs fully self-managed clusters on-premises, and EKS Hybrid Nodes, which connects on-premises worker nodes to an AWS-managed EKS control plane.
What it does well. EKS Anywhere provides operational consistency with cloud EKS by using the same EKS Distro. It supports vSphere, bare metal, and Nutanix, and can operate in air-gapped environments. EKS Hybrid Nodes (launched December 2024) offloads control plane management to AWS, reducing operational burden for connected environments.
Where it falls short. EKS Anywhere requires deep Cluster API (CAPI) expertise. Your team manages the full cluster lifecycle: control plane availability, Kubernetes upgrades, OS patching, node health, and disaster recovery. Bare metal upgrades require spare hardware servers. The Enterprise subscription costs $24,000/cluster/year and requires an AWS Enterprise Support plan starting at $5,000/month. A typical 3-cluster deployment costs at least $114,000/year before infrastructure.
The platform is also shrinking. CloudStack and AWS Snow providers were removed in v0.26, Ubuntu 20.04 was dropped in v0.25, and AWS now steers customers toward EKS Hybrid Nodes for connected environments. EKS Hybrid Nodes requires reliable connectivity to AWS (minimum 100 Mbps, maximum 200ms RTT), runs a mixed CNI architecture that adds complexity, and uses per-vCPU-hour pricing where hyperthreading doubles the bill.
Neither product supports multi-cloud. EKS Anywhere runs only on your own infrastructure, and EKS Hybrid Nodes connects only to AWS.
Best for: Organizations with deep AWS expertise that need on-premises Kubernetes compatible with their existing EKS workflows and are prepared for significant operational responsibility.
3. Cloudfleet
Cloudfleet takes a different architectural approach from the other platforms on this list. Instead of managing separate clusters per environment or adding a management layer on top of existing clusters, Cloudfleet provides a single unified Kubernetes cluster that spans multiple clouds and on-premises locations.
What it does well. Cloudfleet is a fully managed platform. The control plane, node provisioning, networking, upgrades, and scaling are all handled for you. Adding an on-premises node requires a single CLI command on any Linux machine. No Kubernetes expertise is needed to get started. The platform supports 12+ cloud providers (AWS, GCP, Azure, Hetzner, OVHcloud, Exoscale, and others) plus on-premises infrastructure (bare metal, VMware vSphere, Proxmox VE) from a single control plane.
Networking is built in. All nodes connect through an encrypted, WireGuard-based peer-to-peer overlay network regardless of provider or location. DNS-based global load balancing and service exposure work across environments out of the box. Cost optimization features include dynamic node auto-provisioning, automatic spot instance failover across providers, and workload-aware scheduling.
Pricing is transparent. The Basic plan is free for clusters up to 24 vCPUs with no credit card required. The Pro plan is $79/month and includes the managed control plane, networking, monitoring, and support. There are no separate charges for security, policy enforcement, or observability.
As a European-headquartered company, Cloudfleet is not subject to the US CLOUD Act. It supports running workloads exclusively on EU-based infrastructure providers, which matters for organizations with data sovereignty requirements.
Where it falls short. Cloudfleet is a newer platform compared to OpenShift or EKS. Organizations that need deep integration with a specific hyperscaler ecosystem (Azure AD, AWS IAM, GCP BigQuery) may find that native managed Kubernetes services offer tighter integration with those specific cloud services, though Cloudfleet does support Workload Identity Federation for keyless access to cloud provider APIs.
Best for: Organizations that want a fully managed, provider-agnostic Kubernetes platform for hybrid and multi-cloud deployments without the operational overhead of self-managing clusters.
4. Google GKE Enterprise (formerly Anthos)
GKE Enterprise is Google Cloud’s hybrid and multi-cloud Kubernetes management platform. Originally launched as Anthos in 2019, rebranded to GKE Enterprise in 2023, and consolidated into unified GKE in September 2025, it includes fleet management, Config Sync (GitOps), Policy Controller, and Cloud Service Mesh (managed Istio).
What it does well. GKE Enterprise provides a consistent GCP-style management experience for clusters running on Google Cloud, on-premises (via Google Distributed Cloud), and attached third-party clusters. Config Sync and Policy Controller provide strong GitOps and governance capabilities. The service mesh integration is mature and well-documented.
Where it falls short. Google deprecated GKE on AWS and GKE on Azure in March 2025, with full shutdown scheduled for March 2027. The replacement, “attached clusters,” simply registers existing EKS or AKS clusters with GCP fleet management rather than deploying native GKE clusters. This significantly limits the multi-cloud promise.
Pricing charges $0.00822/vCPU/hour for cloud clusters ($6/vCPU/month) and $0.03288/vCPU/hour for on-premises clusters ($24/vCPU/month), plus separate Google Cloud support fees. Historically, Anthos required a $10,000/month minimum commitment. Operating GKE Enterprise effectively requires deep GCP expertise, and the tight integration with Google Cloud services for monitoring, logging, and policy enforcement creates significant platform lock-in.
Market adoption remains limited. Only around 54 companies worldwide were using Anthos GKE on-premises in 2025, and mindshare in the hybrid cloud category dropped from 9.1% to 3.4% by December 2024. The repeated rebranding signals ongoing strategic uncertainty.
Best for: Organizations already deeply invested in Google Cloud that want to extend GCP management to a small number of on-premises or third-party clusters.
5. Red Hat OpenShift
Red Hat OpenShift is the most established hybrid Kubernetes platform, backed by Red Hat (IBM). It provides a comprehensive platform-as-a-service stack on top of Kubernetes, including integrated CI/CD pipelines, a container registry, service mesh, and advanced security features.
What it does well. OpenShift offers the broadest enterprise feature set of any platform on this list. It runs on any major cloud provider and on-premises infrastructure. Advanced Cluster Management (ACM) provides centralized governance across multiple clusters. The developer experience includes built-in CI/CD, source-to-image builds, and a comprehensive web console. The platform has strong adoption in regulated industries (finance, government, healthcare) and a large ecosystem of certified operators.
Where it falls short. OpenShift’s per-core licensing model is the most common complaint. Red Hat switched from per-socket to per-physical-core pricing, and organizations report 300-500% price increases at renewal. A typical mid-size deployment (72 vCPUs) costs $92,000-$280,000/year including infrastructure. Scaling up to more powerful hardware triggers proportionally higher licensing costs, creating what users call a “success penalty.”
Operational complexity is significant. OpenShift introduces proprietary abstractions like Security Context Constraints (SCCs) and DeploymentConfigs that do not exist in standard Kubernetes. The platform bundles 100+ open-source projects, making upgrades a major effort. The control plane and default monitoring stack consume substantial compute resources. For small clusters, a large percentage of compute goes to running OpenShift itself.
Multi-cloud deployments work in theory but break down at the networking and storage layers. Persistent volumes cannot migrate across clouds, and each cloud environment requires different identity providers and networking configurations. Managed variants (ROSA on AWS, ARO on Azure) impose dual billing and restrict access to underlying infrastructure.
Mindshare is declining: from 23.4% to 18.9% in the container management category as of January 2026, as organizations increasingly evaluate lighter, less expensive alternatives.
Best for: Large enterprises that need a comprehensive, opinionated platform-as-a-service and can absorb the licensing costs and operational complexity.
Side-by-side comparison
| Azure Arc | EKS Anywhere | Cloudfleet | GKE Enterprise | OpenShift | |
|---|---|---|---|---|---|
| Type | Management overlay | Self-managed distro | Fully managed platform | Fleet management | Platform-as-a-service |
| Creates clusters | No | Yes (self-managed) | Yes (fully managed) | Yes (GCP only) | Yes (self-managed) |
| Node provisioning | No | No | Yes (auto-provisioning) | Partial (GCP only) | No |
| Multi-cloud | View only | No | Yes (12+ providers) | Limited (attached clusters) | Separate clusters per cloud |
| On-premises | Yes (overlay) | Yes (vSphere, bare metal) | Yes (bare metal, vSphere, Proxmox) | Yes (Google Distributed Cloud) | Yes (any infrastructure) |
| Cross-cloud networking | No | No | Yes (WireGuard overlay) | Partial (service mesh) | Partial (service mesh) |
| Operational burden | High (you manage everything) | Very high (full lifecycle) | Low (fully managed) | Medium (GCP expertise) | High (complex platform) |
| Entry price | Free (6 vCPUs) | Free (OSS, no support) | Free (24 vCPUs) | ~$600/mo (100 vCPUs) | ~$12,000/yr (6 core-pairs) |
| Real production cost | $900-1,700/mo (80 vCPUs) | $114,000+/yr (3 clusters) | $79/mo (Pro tier) | $10,000+/mo (enterprise) | $92,000-280,000/yr |
| Data sovereignty | US CLOUD Act applies | US CLOUD Act applies | European HQ, no CLOUD Act | US CLOUD Act applies | US CLOUD Act applies (IBM) |
Which platform fits which use case
You have existing clusters and want Azure governance. Azure Arc is designed for this. It adds visibility and policy enforcement without requiring infrastructure changes. Just be prepared for the hidden costs of production add-ons.
You are all-in on AWS and need on-premises Kubernetes. EKS Anywhere or EKS Hybrid Nodes keeps you in the AWS ecosystem. Budget for the Enterprise subscription, mandatory support plan, and the operational team to manage cluster lifecycles.
You want a fully managed platform without cloud lock-in. Cloudfleet handles the entire stack and works across all major providers and on-premises infrastructure. The transparent pricing and European headquarters are advantages for organizations that care about cost predictability and data sovereignty.
You are deeply invested in Google Cloud. GKE Enterprise extends GCP management to on-premises and third-party clusters. Be aware that multi-cloud support has narrowed significantly since the Anthos days, and the product continues to evolve through rebranding and feature changes.
You need a comprehensive platform-as-a-service for regulated industries. OpenShift provides the broadest feature set with deep CI/CD integration and strong compliance tooling. Factor in the per-core licensing costs and the operational complexity of managing the platform itself.
There is no single best platform for every organization. The right choice depends on your existing infrastructure, cloud provider relationships, team expertise, budget, and regulatory requirements. If you are evaluating specific platforms against Cloudfleet, we have detailed comparison pages for Azure Arc, EKS Anywhere, GKE Enterprise, and OpenShift.
