The Data Behind the Private Cloud Comeback in 2026

For a long time, the conventional wisdom in infrastructure was simple: public cloud is the future, everything else is legacy, and anyone who disagrees is just resisting change. That narrative made sense when organizations were still discovering what cloud could do for them. It makes less sense when the bills arrive, the audit findings stack up, and the VMware renewal quote lands at four times last year’s number.

The shift happening in 2026 is measurable, documented, and accelerating. Here’s what the data actually shows.

The OpenStack Adoption Numbers

The most striking data point in the current infrastructure landscape comes from the Open Infrastructure Foundation. VMware migrations and public cloud repatriations are set to drive a quadrupling of OpenStack deployments by 2029, according to OIF projections. That’s not a marginal uptick. That’s a structural shift in how organizations are thinking about their infrastructure.

The market size data supports it. Mordor Intelligence estimates the OpenStack market at $22.81 billion in 2024, growing to $91.44 billion by 2029 at a CAGR of 32 percent, driven by OpenStack’s role as a VMware alternative and its suitability for AI and high-performance computing workloads.

The scale of current production deployments is also worth noting. More than 55 million cores are estimated to be in production globally, with mega-users including Walmart, Workday, and CERN operating one million or more cores each. The largest growth segment isn’t hyperscalers or governments. It’s small and medium-sized businesses, which is the segment most directly affected by VMware licensing changes and public cloud cost pressure.

The drivers behind the growth are specific. Over 80% of OpenInfra Foundation members have received requests from organizations to migrate workloads from VMware to OpenStack following Broadcom’s acquisition. Data sovereignty regulations requiring on-premises or private cloud infrastructure are accelerating the trend. And AI workloads, which benefit from the kind of dedicated hardware access that private cloud provides, are pulling organizations toward infrastructure they control rather than infrastructure they share.

The Cloud Repatriation Data

The OIF numbers reflect what broader market research is also finding. Organizations aren’t just evaluating private cloud alternatives theoretically. They’re moving workloads.

Flexera’s 2025 State of the Cloud Report found that 21% of cloud workloads have already been repatriated back to private infrastructure. That’s not a fringe phenomenon. It represents a substantial portion of the workloads that organizations moved to public cloud over the last decade, now moving back because the economics stopped making sense.

The cost pressure driving repatriation is well documented. 84% of organizations cite cost management as their top cloud challenge, and organizations exceed their cloud budgets by an average of 17%. These aren’t one-time oversights. They’re structural features of usage-based billing at scale that consistently produce budget surprises.

The hybrid reality is also clear in the data. 70% of organizations now run hybrid architectures, using at least one public and one private cloud. The move isn’t from public cloud to private cloud wholesale. It’s from single-provider public cloud to a more deliberate architecture that puts different workloads where the economics and control requirements make the most sense.

GEICO is a concrete example of the trend. The company decided to repatriate workloads from public cloud to run on Kubernetes deployed on top of OpenStack, a move that reflects the pattern showing up across the industry: mature organizations reassessing which workloads genuinely benefit from public cloud elasticity versus which ones are paying a premium for flexibility they don’t use.

The VMware Displacement Numbers

The cloud repatriation trend is significant on its own. Layered on top of it is the VMware displacement event, which is accelerating the timeline considerably.

Broadcom’s acquisition of VMware in November 2023 triggered changes that have been working their way through enterprise renewal cycles ever since. The specifics are by now well known but worth stating concisely: perpetual licensing is gone, the product catalog consolidated from 168 products to four bundles, and per-core licensing with a 72-core minimum means organizations running smaller deployments pay for capacity they’ll never use. Renewal cost increases ranging from 150% to over 1,500% have been reported across European and US customers.

Gartner predicts VMware will lose 35% of workloads within two years of the licensing changes. Given that VMware previously held roughly 40% of the virtualization market, that represents a migration event affecting a substantial portion of enterprise infrastructure globally.

The migration window matters. Most organizations facing VMware renewal decisions will make their platform choices within the next 12 to 18 months. The organizations that evaluate alternatives carefully and move deliberately are in a better position than those who defer the decision until a renewal deadline forces it.

Where the Workloads Are Landing

The data on where displaced VMware workloads and repatriated cloud workloads are landing points in a few consistent directions.

OpenStack is the primary destination for organizations that need full private cloud capabilities: multi-tenant environments, API-driven provisioning, Kubernetes integration, and the ability to run infrastructure-as-code workflows that translate from public cloud environments. The OIF’s 80%+ member survey finding reflects real demand, not projection.

Proxmox has emerged as the grassroots favorite for VMware administrators specifically. Its web interface maps closely to what vSphere users already know, VM import tooling handles OVA and OVF formats directly, and benchmarks have shown Proxmox beating VMware ESXi in 56 of 57 performance tests with IOPS gains approaching 50%. For organizations that want a faster migration path without the learning curve of a full cloud platform, Proxmox is where many teams are landing.

The bare metal cloud market is growing to serve both groups. The global bare metal cloud market is forecast to grow from $14.32 billion in 2025 to $36.71 billion by 2030, reflecting sustained demand for dedicated infrastructure that provides the control and performance of physical hardware with the operational model of cloud services.

OpenMetal’s hosted private cloud sits in the middle of these trends: OpenStack on dedicated hardware for organizations that want private cloud capabilities without building a data center, and bare metal servers for teams that want to run Proxmox or other hypervisors on hardware they control directly.

What the Data Doesn’t Tell You

The repatriation trend is real, but the data doesn’t support a wholesale abandonment of public cloud, and it’s worth being clear about that.

Flexera’s finding that 70% of organizations run hybrid architectures is the more instructive data point than the 21% repatriation figure. Most organizations aren’t moving everything off public cloud. They’re moving specific workloads: steady-state production infrastructure where fixed costs beat variable billing, compliance-sensitive environments where dedicated hardware satisfies audit requirements that shared cloud can’t, and high-performance workloads where bare metal delivers better performance than virtualized cloud instances.

The workloads that stay on public cloud are also well understood. Genuinely elastic workloads with unpredictable demand still benefit from pay-as-you-go pricing. Development environments that spin up and tear down frequently. Managed services like machine learning platforms and sophisticated analytics engines that would be expensive to replicate. The data doesn’t say public cloud is wrong. It says that running everything on public cloud, regardless of workload characteristics, is wrong.

The organizations that are navigating this well are the ones that have done honest workload analysis rather than making sweeping platform decisions. That analysis tends to produce hybrid architectures with clear criteria for which workloads belong where, rather than migrations driven by cost panic or vendor pressure.

What This Means If You Haven’t Evaluated Recently

If you last looked seriously at private cloud or VMware alternatives two or three years ago, the market has changed enough that the evaluation is worth revisiting.

The OpenStack of 2026 is operationally different from the OpenStack of 2021. Hosted deployments that arrive preconfigured and production-ready in minutes, managed operations that absorb the infrastructure complexity, and tooling that integrates with the same Terraform and Ansible workflows that run on public cloud have addressed the operational complexity concerns that kept many organizations away. The Proxmox option for VMware administrators has also matured significantly, with Proxmox VE 9.0 built on Debian 13 and actively developed.

The economics have also shifted in ways that weren’t true two years ago. The hardware supply crisis has raised public cloud costs alongside private cloud hardware costs, but fixed-cost dedicated infrastructure means your bill doesn’t move after you’ve committed, which is a different proposition from variable cloud billing during a period of hardware cost inflation.

The cloud deployment calculator gives you itemized private cloud pricing without a sales call, which makes a first-pass comparison against your current infrastructure spend straightforward. The migration resources on OpenMetal’s site cover the practical migration paths for both VMware and public cloud workloads.

The trend is clear. The question for any specific organization is which workloads belong in the movement and which ones don’t. That analysis is easier to do now than it’s ever been.

Interested in understanding how your specific workloads compare? Use the cloud deployment calculator or contact the OpenMetal team to discuss your infrastructure situation.