Multicloud Container Platforms

Kubernetes is open source, standardised, and free. Every major cloud provider offers a managed version. Anyone can install it.

So what exactly does a multicloud container platform sell?

The answer is not Kubernetes. It is everything around running a lot of Kubernetes, in places that differ, for years, without a team of specialists per cluster. Forrester's framing of the category describes platforms enabling the development, operations, and management of containerised applications across public clouds, edge, and on-premises environments.

The operative word is across. One cluster is an engineering exercise. Four hundred clusters spanning three clouds, a data centre, and two hundred retail sites is a fleet management problem, and fleet management is the product.

What the fleet problem actually involves

Running Kubernetes at scale produces a set of recurring difficulties that have nothing to do with container orchestration itself.

Version currency is the first. Kubernetes releases frequently and support windows are short, which means an organisation running many clusters is perpetually upgrading. Forrester's assessment of Nutanix credits cluster lifecycle operations specifically, singling out seamless non-disruptive upgrades, which tells you how much that matters in practice.

Consistency is the second. A cluster on one cloud, another on a different cloud, and a third in a data centre will drift in configuration, policy, networking, and storage behaviour unless something enforces sameness. Drift is where the multicloud promise fails, because an application that runs on one and not the others is not portable in any useful sense.

Governance is the third. Who can deploy what, where, with which permissions, subject to which policies, and how is that demonstrated to an auditor across a fleet.

And support is the fourth, which sounds unglamorous and is frequently the purchase justification. Red Hat's citation includes service level agreements of 99.95 percent for its public cloud managed versions, with Forrester noting this showcases the ability to engineer capabilities beyond those of native public cloud services.

That is the honest answer to the opening question. You are buying somebody's commitment to keep the fleet running, expressed contractually.

Inside The Forrester Wave: Multicloud Container Platforms, Q3 2025

The evaluation scored nine providers against thirty one criteria across current offering, strategy, and market presence, following a Landscape published in Q1 2025.

Red Hat placed as a Leader with the highest possible score across twenty seven of the thirty one criteria and the highest score in both current offering and strategy. Forrester credited OpenShift with excelling in core Kubernetes areas including operator options, management, GitOps automation, and flexible interfaces, and noted that strong execution has kept it among the top players.

Nutanix placed as a Leader with maximum scores in cluster operations, persistent storage, and edge computing. Forrester noted its platform builds on capability acquired from D2iQ, and positioned it for companies seeking a cloud-native complement to migration away from VMware as well as for intermittently connected and air-gapped edge scenarios.

SUSE also placed as a Leader.

Spectro Cloud placed as a Strong Performer, with Forrester observing that its current offering is on or above par across the board with no obvious gaps, which it described as notable for a startup, and positioning it for enterprises balancing cloud-native modernisation and AI enablement with continued support for legacy virtual machines.

Kubermatic was positioned for companies seeking highly automated, scalable infrastructure on which to base a larger platform, appealing to cost-conscious leaders who value simplicity over platform bloat, particularly advanced Kubernetes users.

Mirantis, Canonical, Diamanti, and Rafay completed the field.

A market reshaped by someone else's licensing

The most consequential force acting on this category comes from a vendor that is not in it.

Nutanix's positioning statement names it directly: a cloud-native complement to migration away from VMware. Reporting on the evaluation noted that VMware was repeatedly framed in the research as a platform enterprises might want to move away from.

Following Broadcom's acquisition of VMware, licensing and packaging changes prompted a substantial population of enterprises to reassess a virtualisation platform many had run without question for fifteen years. Those organisations are not simply choosing a replacement hypervisor. A meaningful proportion are asking whether the workloads should be virtual machines at all.

That produces an unusual commercial opportunity for container platform vendors. An organisation forced to revisit its infrastructure platform is, for the first time in years, genuinely open to a different architecture.

It also explains why several vendors in this evaluation emphasise virtual machine support alongside containers. Spectro Cloud's positioning names support for legacy VMs explicitly. A platform that can run both, on the same infrastructure, under the same management, lets an organisation migrate off VMware without simultaneously rewriting its applications.

That is a considerably easier proposition than container-only migration, and it is the reason this market is more contested now than it was three years ago.

The last-mile problem

The second force is edge, and Forrester frames it precisely: continued innovation at the edge addressing automation, management, intermittently disconnected operations, and AI inferencing, responding to the last-mile problem created by large region-based public cloud infrastructures.

Hyperscale cloud is built around a small number of very large regional facilities. That architecture is excellent for centralised workloads and structurally wrong for anything that must run where the physical world is.

A factory floor cannot depend on a link to a data centre several hundred kilometres away for a control decision. A retail store needs to keep transacting when connectivity drops. A ship, a mine, a wind farm, or a military deployment may be disconnected for extended periods by design.

Kubernetes at the edge therefore has requirements that centralised Kubernetes does not. Small footprint, because there is no rack. Air-gapped and intermittently connected operation, because there may be no network. Remote lifecycle management, because there is nobody on site to fix it. And resilience to being powered off without warning.

Nutanix's maximum score in edge computing and Forrester's note about air-gapped and intermittently connected environments describe exactly this capability set, and it is the clearest area where these platforms do something the hyperscalers' native services do not.

For a European organisation there is a further dimension. Edge deployment keeps data physically local by design, which intersects with data residency requirements in a way that is architecturally clean rather than contractually asserted.

Comprehensive against minimal

Kubermatic's positioning contains the market's real split, stated more bluntly than most analyst material manages: cost-conscious leaders who value simplicity over platform bloat.

That names a genuine disagreement about what a container platform should be.

The comprehensive position is that Kubernetes alone is insufficient for enterprise use. It needs an integrated registry, build pipelines, service mesh, observability, security scanning, policy enforcement, developer portals, and a supported set of operators, all tested together and supported as one thing. Red Hat's twenty seven maximum scores describe a platform built on that premise.

The minimal position is that Kubernetes is the standard, everything else should be composable, and a platform that bundles opinionated choices is selling you decisions you may not want alongside licence cost you certainly do not.

Both are defensible and they suit different organisations.

An enterprise without deep platform engineering capability benefits enormously from the comprehensive option, because assembling and maintaining that stack yourself is a permanent team. An organisation with strong platform engineers frequently finds the bundled components inferior to what they would choose, and resents paying for them.

The practical test is whether your platform team has opinions. If they do, and those opinions are informed, a minimal foundation plus their choices will outperform a bundle. If they do not, the bundle is buying you a set of decisions made by people who do this full time.

Vertical stacks and AI

Forrester's Landscape research observed that Kubernetes-based platforms increasingly feature integrations branching into discrete vertical stacks, including generative AI, machine learning, internet of things, 5G, and other domains, all on a unified control plane.

That is a description of Kubernetes' actual role, which is less container orchestration and more general-purpose infrastructure abstraction. It schedules workloads onto resources with declared requirements, and it does not much care whether the workload is a web service, a model training job, or a network function.

AI made this visible. Training and inference need GPUs, and GPUs need scheduling, sharing, and allocation policy across teams competing for scarce hardware. Kubernetes already does scheduling, so the AI infrastructure stack largely built itself on top rather than beside it.

Inferencing at the edge is where the two forces in this article converge. A model running in a factory, a store, or a vehicle needs the edge characteristics described above and the GPU scheduling characteristics described here, on the same platform, managed centrally.

That combination is where these vendors are competing hardest, and it is a genuinely new requirement rather than a repackaging of existing capability.

What this leaves a buyer

Three questions decide more than the tier placement.

How many clusters will you actually run, and where. A handful in one cloud does not justify this category, and the cloud provider's managed service will serve you. The economics change sharply somewhere above a few dozen clusters across environments, and again once edge sites are involved.

Whether you need virtual machines alongside containers. For organisations working through VMware alternatives this is the decisive capability, and it separates the field.

And whether your platform team has informed opinions about the stack. That determines whether comprehensive or minimal is the right shape, and it is a question about your organisation rather than about the vendors.

The underlying technology is standardised and free, which is unusual and worth remembering during a procurement. What you are buying is somebody else's operational discipline, expressed as upgrades that do not break things and a support number that answers. Those are real and they are worth paying for. They are also not the same thing as capability, and evaluations that score capability will not surface them.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.