The News
Kubermatic is seeing accelerating demand across two converging trends: organizations migrating away from VMware toward Kubernetes-based virtualization platforms and enterprises shifting AI inference workload management from data science teams back to traditional IT operations and platform engineering groups. At KubeCon NA 2026, Kubermatic described a third major demand driver, particularly pronounced in Europe: digital sovereignty and cloud repatriation, as organizations move compute back on-premises to satisfy regulatory requirements, reduce cloud spend, and keep AI inference close to the data it processes.
Analyst Take
The VMware Exodus Is Becoming a Platform Consolidation Play
The VMware migration story has been told many times over the past two years, but this conversation reveals something more interesting than simple hypervisor substitution. Organizations aren’t just replacing VMware with another VM platform. The smarter ones are using the forced migration as the trigger to consolidate what had become three separate operational silos: a Kubernetes platform team, a VMware team, and an emerging AI infrastructure team. That’s three separate toolchains, three separate on-call rotations, and three separate talent pools to maintain. The business case for collapsing all three onto a single Kubernetes-based platform is straightforward, and the operational cost argument is compelling to any CIO managing headcount pressure.
For developers and platform engineers, the architectural implication is significant. KubeVirt allows VMs to run as first-class Kubernetes objects, meaning existing VM-based workloads don’t require immediate refactoring to move onto the platform. That’s a pragmatic on-ramp. Kubermatic was explicit that customers can start with VMs today and containerize incrementally, without a platform migration mid-journey. That staged approach addresses one of the most common objections to Kubernetes adoption among infrastructure-heavy shops: the refactoring cost of lifting existing workloads.
AI Infrastructure Is Drifting Back to IT Ops
The observation about AI workload ownership deserves more attention than it typically gets. Early AI deployments were largely self-provisioned by data science teams, running Docker on bare metal or ad-hoc VMs. That model is breaking down. Data scientists are not infrastructure operators, and organizations are discovering that GPU clusters require the same operational discipline as any production infrastructure: patching, monitoring, capacity planning, and incident response. The natural owner is the platform or IT operations team.
This shift has a direct procurement implication. AI infrastructure budgets, which were previously held by ML teams and spent on ad-hoc compute, are increasingly flowing through the same IT operations channels that procure platform software. That changes the buying motion significantly: the decision-maker is now a platform architect or VP of infrastructure, not a chief data scientist or AI product lead. Vendors selling into this space need to position accordingly.
Sovereignty and Repatriation Are Real, But Geography-Dependent
The sovereignty conversation is genuine in Europe and slowly catching up in North America, at least for now. The EU Cyber Resilience Act, GDPR data residency requirements, and growing political pressure to reduce dependence on U.S. hyperscalers are producing concrete procurement decisions. Kubermatic’s German headquarters is a legitimate differentiator in that context, not just a marketing angle. Insurance companies required to keep policyholder data within specific national borders, manufacturers running AI inference on shop-floor data that can’t be streamed to a public cloud endpoint, healthcare organizations processing medical imaging at the point of care: these are real use cases driving on-premises Kubernetes adoption in ways that the U.S. market hasn’t yet fully absorbed.
The repatriation dynamic also intersects with AI economics. Moving tokens to a public cloud endpoint and back is expensive when inference volumes scale. Processing data where it lives, whether that’s a factory floor, a hospital imaging system, or a financial services data warehouse, is increasingly the operationally and economically rational choice. According to ECI Research’s 2026 Application Development: Day 2 survey, 49.0% of respondents identified “Data volume growth” as one of their biggest observability challenges. That same data gravity problem applies to inference: when datasets become too large to move economically, the compute has to come to the data.
The multi-cloud skills gap compounds all of this. Running two or more cloud environments is the norm, not the exception, and finding engineers with deep operational knowledge across multiple hyperscalers is genuinely difficult. ECI Research’s 2026 Application Development: Day 2 survey found that 42.4% of respondents are running between 6 and 10 observability tools simultaneously, a number that reflects the broader fragmentation of the infrastructure management stack. A platform that presents a consistent operational model across AWS, Azure, GCP, and on-premises bare metal addresses a real pain point, not a theoretical one.
Looking Ahead
The VMware displacement cycle still has significant runway. Many organizations have delayed migration decisions while evaluating options, and the combination of Broadcom’s pricing changes and the maturing KubeVirt ecosystem is pushing more of those fence-sitters to act in 2027. Vendors that can offer a credible path from pure VM management to container-native workloads to AI inference on a single control plane are positioned to capture a disproportionate share of that migration spend. The competitive battle between OpenShift and Rancher/SUSE will intensify, and smaller platform-focused vendors will need to compete on operational simplicity and total cost of ownership rather than feature breadth.
The sovereignty and repatriation trend is also likely to intensify, driven by regulatory tightening in the EU and growing enterprise interest in AI cost management. As inference workloads scale and the economics of public cloud GPU compute become harder to justify for high-volume, low-latency use cases, the on-premises Kubernetes platform argument gets stronger. Organizations that invest now in a unified operational model spanning cloud and on-premises environments will have a meaningful structural advantage when AI workloads mature from experimental to mission-critical. The platform teams building that foundation today are making a decision that will shape their infrastructure costs and operational agility for the next decade.
Stay Ahead of Application Development Trends
Get weekly analyst insights, research notes, event coverage, and AppDevANGLE updates delivered directly to your inbox.
Subscribe for Weekly Insights
Join technology leaders, practitioners, and GTM teams following the trends shaping modern software delivery.
Looking for deeper research access?
Explore ECI Research reports, survey insights, and market analysis through the ECI Research Portal.
