Bundling Is Not Unifying. Multicluster Kubernetes Management Still Has a Gap.

Vendors keep folding more products into one bill. Buyers read a single line item as a single platform. Those are different things, and the gap between them is where multicluster Kubernetes management gets hard. Red Hat’s 2026 fleet management report puts numbers on it. Eighty-five percent of organizations grew their cluster count in the past year. Seventy percent run Kubernetes across more than one public cloud. Forty-three percent say the complexity of managing that fleet went up. Growth is the easy part. Coherence is not.

Two clusters can ship in the same suite and still run different versions, different network policies, and different identity rules. A bundle cleans up procurement. It does not clean up operations unless the parts were built to share state.

One Invoice Is Not One Control Plane

SUSE Rancher Prime manages upstream Kubernetes, the CNCF graduated project, on any conformant distribution. That covers on prem and every major cloud. The Rancher control plane, Fleet for GitOps across a large fleet, and the distribution itself are separate projects on separate release cycles. Portability is the win. Integration is the bill you pay for it, and you assemble the operational story yourself.

VMware Tanzu and VMware Cloud Foundation (VCF after first mention) push the other way. Tanzu is built for VCF, and VCF is the private cloud underneath it. VCF 9.1 lifted the ceiling to 500 Kubernetes clusters per Supervisor. For a private cloud, that is serious scale. The catch is direction. Workloads that must live on AWS or Azure are guests in a house built for vSphere.

Red Hat OpenShift sits between the two. Advanced Cluster Management gives one hub that creates and governs OpenShift clusters on AWS, Azure, Google Cloud, and bare metal. It imports other distributions too. Imported clusters get a thinner lifecycle than native ones. Full create, upgrade, and destroy works best when the cluster is OpenShift. The control plane is genuinely unified. Depth still depends on which Kubernetes you picked.

Open source shows how wide the gap runs. Karmada, a CNCF incubating project, treats many clusters as one scheduling target across clouds. Crossplane, now CNCF graduated, does the same job for infrastructure through Kubernetes APIs. Neither ships a vendor console. Both exist because the bundled consoles stop at the boundary of what the vendor sells.

What Unifying Actually Requires

Ask a suite four questions. Does one inventory list every cluster and its version? Does one identity provider issue every workload credential? Does one policy engine block a bad deployment before it lands rather than after? Does one ledger show which team spent what? A vendor that answers yes to three has bundling with good marketing. A vendor that answers yes to all four has a platform.

Drift is the tax you pay without a control plane. An AWS cluster picks up a new storage driver. An Azure cluster runs last quarter’s ingress controller. Nobody notices until a deploy fails in one region and works in another. That is not a product gap. It is an ownership gap, and suites do not fix ownership.

Cost is the question most suites dodge. A cluster nobody owns still bills every month. Kubernetes does not track spending on its own. Fleet dashboards rarely roll cost up by namespace or by team, which is the only granularity finance actually cares about. That gap leaks budget, and no bundle closes it for free.

The multi-provider reality is not going away either. Most fleets already span more than one cloud, and air-gapped sites widen the problem. A control plane that sees one substrate leaves the rest to scripts and tribal knowledge. Vendors know this. It is why every suite claims portability and why so few prove it.

Unifying also changes who you hire. One control plane across four providers is one skill set to staff and one policy to audit. Four consoles stitched together by scripts is four skill sets and an on-call rotation that guesses which tool owns the truth. Those savings show up in audit time and headcount, not on the license line.

Buy the Control Plane, Not the Box

The honest test runs two weeks. Put three clusters on three providers. Import them into the candidate platform. Then change one policy everywhere at once and prove it took effect. Vendors that pass earn the license. Vendors that hand you a nicer console and a longer feature list are selling bundling. That is fine, as long as you price it as bundling.

Platform engineering is one of the forces in the 2026 State of Enterprise Infrastructure report, which covers how the operating layer changed and why maturity tracks with outcomes.

Related reading. A Full VMware Exit Costs More Than Three Years of Broadcom’s Bill. Karmada Just Graduated. Multi-Cluster Kubernetes Is Now a Solved Problem, Sort Of.. Digital Sovereignty Is a Procurement Problem, Not a Data Center Problem. The 2026 State of Enterprise Infrastructure.