The VMware exodus is real, but partial.

The phrase "VMware customer exodus" appeared in trade press within months of the Broadcom acquisition closing, and it has been a contested phrase ever since. Critics say the exodus is exaggerated marketing from competitors; defenders point to a steady drumbeat of customer announcements and migration projects. The truth is more measurable than the headlines suggest. There is movement; it has a shape; and the shape matters for any IT team trying to make a rational platform decision.

This article documents the exodus pattern as we observe it in 2026 — which workloads are leaving first, which alternatives are receiving them, the timing curves, and the segments where movement is essentially zero. It is built from advisory engagement data, public customer migration announcements, and competitor disclosures.

What 'exodus' actually means in this market

A useful way to think about platform movement is to separate three different decisions enterprises are making.

The first is the no-new-VMware decision: a quiet but durable choice that new workloads will not land on vSphere by default. This is the most common pattern we see, and it is rarely announced publicly. It does not generate near-term revenue movement, but it shapes the install-base curve over three to five years.

The second is the workload migration decision: actively moving existing VMs off vSphere onto an alternative platform. This is much rarer than the first and is concentrated in specific segments. When it happens, it is usually phased over 18–36 months, and it tends to start with the easiest workloads (stateless, simpler applications) and end with the hardest.

The third is the full platform exit: a decision to remove VMware entirely. This is the rarest of the three at enterprise scale. It is usually only feasible for organisations with a small enough VMware footprint, a strong enough engineering team, and a long enough planning horizon to absorb the migration cost.

When trade press talks about an exodus, it is usually conflating these three. The first is widespread; the second is real but segmented; the third is rare.

Which workloads are moving first

Across the engagements where customers are actively moving workloads off vSphere, four categories show up repeatedly as the early migrators.

Test and development environments

Test/dev is the most common early-exit workload. The compliance bar is low, the application owners are typically more flexible, and the per-VM cost differential becomes painful first in environments that scale up and down often. We see test/dev migrations onto Proxmox, Hyper-V, hyperscaler-native compute, and increasingly Kubernetes-based platforms.

VDI and remote desktop workloads

Horizon VDI has been particularly exposed under Broadcom's pricing changes, and the alternatives — Citrix DaaS, Windows 365, Azure Virtual Desktop — have become commercially competitive. VDI is also operationally separable from the rest of the estate, which makes migration tractable.

Edge and ROBO sites

Small remote-office deployments — three or four hosts per location — are economic outliers under VCF pricing. Many enterprises are quietly replacing these with Hyper-V (covered under existing Windows entitlement), hyperscaler thin clients, or single-vendor HCI alternatives. The migration is invisible at headline level but adds up across hundreds of sites.

Greenfield application platforms

Net-new application platforms — particularly cloud-native and Kubernetes-anchored — are increasingly landing outside vSphere. This is not, strictly, an exodus; it is a non-arrival. But it has the same effect on the long-term install-base curve.

Where the movement is going

Nutanix

Nutanix has been the most visible commercial beneficiary. Their AHV hypervisor, packaged with HCI and a coherent operational model, gives mid-market and lower-enterprise customers a credible single-vendor alternative. Their public new-customer numbers correlate with what we see in advisory engagements: a steady stream of 100-to-1,000-host customers picking Nutanix as the destination, with 24-to-36-month phased migrations.

Proxmox VE

For technically sophisticated organisations and the smaller end of the mid-market, Proxmox has emerged as the most cited alternative. It is free at the licensing layer, mature enough for production, and increasingly supported by a growing partner ecosystem.

Hyper-V

Microsoft Hyper-V has had a quiet resurgence in Windows-heavy estates where Windows Server datacenter licensing already covers the hypervisor. Microsoft has been deliberate about not marketing Hyper-V aggressively — they would rather customers move to Azure — but the deployment data shows real share gain at the mid-market and edge tiers.

OpenShift Virtualization and KVM-based stacks

For organisations standardising on Kubernetes — particularly those already running Red Hat OpenShift — OpenShift Virtualization gives a path to consolidating VMs and containers on a single platform. Adoption is concentrated but real.

Hyperscaler-native compute

For workloads that can be modernised rather than migrated, the destination is often hyperscaler-native: EC2, Azure VMs, GCE, plus the managed Kubernetes services on top. This is the exodus path with the most strategic upside and the highest engineering cost.

Where the exodus is not happening

Top-tier enterprise data centres

The Fortune 100 VMware estates — multi-thousand-host footprints with deep NSX, vSAN, and Aria integration — are not moving. The migration cost at that scale exceeds three to five years of Broadcom subscription pricing, the operational risk is significant, and the application portfolio is too diverse for any single alternative to absorb. These customers are negotiating hard and tightening their estate, but they are not exiting.

Regulated and government workloads

The recertification burden in these environments makes hypervisor changes very slow. Some workloads are migrating, but the segment-level movement is small.

Highly NSX-dependent estates

Customers with deep NSX investment — particularly micro-segmentation, advanced load balancing, and service-mesh integration — have the highest migration friction of any segment. Alternatives exist, but the operational and security model has to be rebuilt.

Timing patterns we observe

The migration timing curve is consistent enough across engagements to describe in general terms.

Months 0–6: decision made, alternative selected, proof-of-concept built. This phase often follows an audit notice or a renewal proposal that crystallises the strategic question.

Months 6–18: easy workloads migrated. Test/dev, simple greenfield, low-coupling applications.

Months 18–36: medium workloads. Departmental applications, internal-facing platforms, some production workloads with manageable dependencies.

Months 36+: hard workloads or remainder. Tightly coupled production, legacy systems, anything dependent on advanced NSX features.

Most enterprises stall somewhere between phase two and phase three. The hardest 20–30% of workloads tend to stay on VMware permanently, even when the strategic decision was to exit, because the business case for migrating them never quite closes.

Practical migration economics

Translating the strategic question into numbers is where most exodus discussions get vague. The economics are more concrete than the headlines suggest, and the variables that matter most are the same across most enterprises.

The 5-year total-cost framework

The right unit of analysis for a migration decision is total cost of ownership over 5 years, broken into three buckets:

Stay-cost: continued Broadcom subscription at projected pricing, plus internal operations cost, plus expected audit-related cost. For a typical enterprise running VCF subscription at full pricing, this comes out to roughly 60-75% of the 5-year TCO in licensing, 20-30% in operations, and the rest in audit/compliance contingency.

Move-cost: alternative-platform licensing or subscription, plus migration project cost (typically 30-50% of the alternative platform's 5-year licensing for a phased migration; higher for compressed timelines), plus operational ramp cost (training, tooling, process redesign), plus residual-VMware cost (almost every migration leaves some workloads behind).

Hybrid-cost: the most common actual outcome, where roughly 40-60% of workloads move and the rest stay. This typically costs more than a clean full move but less than a stay, and produces better strategic optionality than either.

The math is sensitive to a few key variables: the size of the residual VMware estate after migration, the negotiated subscription rate on that residual estate, and the depth of NSX or vSAN dependencies in the migrating workloads. Enterprises that have run this analysis carefully tend to find that the hybrid path is the most economically defensible for most workload portfolios.

The internal politics of the exodus decision

Migration decisions are not just economic; they are organisationally complex. The internal politics of choosing to move (or not move) off VMware are a meaningful drag on the rate of actual movement.

Three internal dynamics consistently slow migrations even when the economics point clearly:

Operations team identity. Many IT operations teams have built their professional identity around VMware skills. The proposal to change platforms is heard as a proposal to obsolete the team. The right organisational response — investing in the operations team's capability on the new platform, treating the migration as a career development opportunity — is rarely well-executed.

Application owner resistance. Application teams often resist platform changes because they do not see the upside (the change happens beneath their layer) but do see the downside (they have to test and revalidate). The right response is to make the migration largely transparent at the application layer, but that requires deliberate engineering effort.

Security and compliance friction. Every platform change triggers a recertification cycle in regulated environments. Where the security and compliance functions are not aligned with the migration timeline, the project slips repeatedly.

Enterprises that successfully execute migrations tend to have strong central programme management, executive sponsorship at C-level, and explicit airtime for these internal dynamics in the project governance.

How to evaluate alternative platforms

For enterprises in the early stages of considering an alternative platform, a structured evaluation matters more than the specific choice. The questions that matter most:

How does the alternative platform handle the workloads that matter most to your business? Most platforms work well for simple workloads; the differentiators show up in the harder ones.

What does the operational model look like at scale? Demos and POCs frequently miss the difference between running 50 hosts and running 5,000.

What is the vendor's commercial trajectory? You do not want to migrate from one vendor undergoing a difficult transition to another that may be next.

What is the ecosystem of integrations, partners, and skilled labour? A platform with a thin ecosystem can still be a defensible choice, but the operational cost is higher.

What is the realistic exit path off the alternative platform if it does not work out? You are negotiating one platform contract while creating the next one. The exit posture matters.

Signals to track over the next 12 months

Several public signals will materially shape the exodus picture through the back half of 2026 and into 2027:

Broadcom's reported revenue trajectory and customer count. These metrics will either validate or challenge the current strategy on Wall Street's terms.

Major customer migration announcements. A small number of high-profile exits would materially shift internal narrative in other large enterprises.

Alternative-platform funding and product release cadence. The pace at which Nutanix, Proxmox commercial offerings, and KVM-based stacks mature will determine the size of the addressable migration market.

Broadcom pricing-policy changes. If Broadcom moderates its pricing posture, some of the migration momentum will reverse. If it tightens further, more will accelerate.

Audit settlement patterns. If settlement outcomes harden materially, more enterprises will treat migration as the cheaper long-run option. If settlements remain negotiable, fewer will.

The aggregate read across these signals will tell you, by the end of 2026, whether the partial exodus has become a more decisive movement or stabilised at its current scale. Most enterprises do not need to make a final decision today; they do need to make the conditional decisions and start the optionality work that lets them act decisively when the picture clarifies.

The communication challenge with executive leadership

One of the under-discussed aspects of the exodus question is how it is communicated to executive leadership. CIOs presenting a "maybe migrate VMware" position to a CEO or CFO face a difficult communication challenge: the technical reality is nuanced, the timeline is multi-year, and the cost picture is sensitive to many variables.

Effective executive communication on this question shares three characteristics.

Workload-segmented rather than platform-binary. The right framing is not "should we leave VMware" but "which workloads belong where". Executive audiences absorb this framing more easily than the binary one, and it leads to more durable decisions.

Anchored on specific numbers rather than directional argument. 5-year TCO across the relevant scenarios, with explicit assumptions and sensitivities, lands better with finance-trained executives than qualitative pros-and-cons.

Honest about uncertainty. Executives respect IT leaders who acknowledge what they do and do not know. Pretending to certainty on questions that genuinely have multi-year answer windows erodes credibility when reality diverges.

The advisor question

Many enterprises ask whether they need external advisory support to navigate the exodus decision. The answer depends on the size of the estate and the complexity of the workload portfolio, but a few patterns hold.

For enterprises with under 100 VMware hosts, internal capability is often sufficient. The decisions are simpler, the alternatives are clearer, and the migration mechanics are more tractable.

For enterprises with 100 to 1,000 hosts, external advisory support typically produces meaningful incremental value, particularly around negotiation, audit defence, and alternative-platform evaluation.

For enterprises with over 1,000 hosts, external advisory support is essentially required to navigate the question well. The complexity is too high, the dollar amounts are too large, and the strategic stakes are too significant for the work to be done well without specialist input.

The role of timing windows

For enterprises actively planning a partial exit, the timing of when to commit is itself a significant decision. Migrating before a renewal locks in flexibility; migrating after a renewal locks in the renewal economics. We see customers who plan their migration sequencing against their renewal calendar consistently achieve better outcomes than customers who treat the two as separate workstreams. Building the migration timeline backward from the renewal date — rather than letting the migration timeline drift while the renewal arrives on a fixed calendar — produces materially better commercial outcomes and reduces the risk of negotiating a renewal while uncertain about the destination architecture.

What this means for planning

The right read for IT leaders is that the exodus is real but partial. Net-new workloads have meaningfully shifted away from VMware as a default, and certain segments — mid-market, edge, VDI — are seeing material customer movement. But VMware retains very strong installed-base economics at the top of the enterprise market, and most migration projects end with a hybrid estate rather than a clean exit.

The strategic question is not "should we leave VMware?" The strategic question is: which workloads belong on VMware long term, given current pricing, and which do not? Answering that requires a workload-by-workload economic and operational analysis, not a platform-level decision.

For most enterprises, the right plan in 2026 is a controlled, segmented migration of the workloads where the math is clear, combined with hard negotiation and audit defence to protect the long-tail VMware footprint that is not going anywhere. That dual-track posture is what we see succeed in practice. A binary "stay or leave" decision is almost always a more expensive answer than the segmented one.