Leaving VMware in 2026: Why and How to Migrate Your Estate to the Public Cloud

Broadcom closed its acquisition of VMware on November 22, 2023. Almost three years on, most vSphere customers have been through at least one renewal under the new model, and for many it turned VMware into a board-level topic.

This post is for CIOs, cloud architects and migration leads, especially in regulated EU organizations, who must decide what to do with a VMware estate before the next renewal. It covers where things stand in October 2026, when public cloud makes sense and when it doesn’t, how to build a TCO that survives a finance review, and a phased migration approach.

The examples are Azure-first, with AWS and Google Cloud alongside. For the wider program view, see the Cloud Migration Handbook.

Key takeaways

  • VMware is now subscription-only, bundled and licensed per core. The documented floor is 16 cores per CPU, but minimum order sizes and which bundles you can buy vary by country. Get the rules for your contract in writing.
  • Since November 1, 2025, every hyperscaler VMware service requires a portable VCF subscription bought from Broadcom. “VMware in the cloud” gets you out of your datacenter, but Broadcom stays in your cost base.
  • vSphere 8 general support ends on October 11, 2027. Together with the three-year contracts signed in 2024 and 2025, that makes 2026–2027 the decision window for most estates.
  • Cloud isn’t automatically cheaper. A credible TCO covers people, egress, commitment discounts and Microsoft licensing. Free Extended Security Updates in Azure, a big lever for Windows Server 2012, can’t be assumed for Windows Server 2016 or SQL Server 2016.
  • Most estates end up with a mix: VMware-as-a-service as a time-boxed bridge, rehosting to native IaaS for most workloads, and refactoring where the business case is clear.
  • Discovery, dependency mapping and landing zone readiness set the schedule far more than the replication tooling does.

Why: what changed after Broadcom bought VMware

Subscription only, bundled, per core

On December 11, 2023, Broadcom stopped selling perpetual VMware licenses and new support contracts for them. It also cut the portfolio down to two main bundles: VMware Cloud Foundation (VCF) and VMware vSphere Foundation (VVF).

Both bundles are licensed per physical CPU core, with a floor of 16 cores per CPU, so an 8-core processor is billed as 16. VCF includes 1 TiB of vSAN per licensed core and VVF 0.25 TiB. vSphere 9 ships only inside the bundles.

Perpetual licenses keep running, but without support they get no patches. In May 2025, holders whose support had lapsed reported letters from Broadcom asking them to remove patches applied after expiry.

What you can buy depends on where you are. In December 2025 Broadcom confirmed VVF was no longer available in some EMEA countries, pushing smaller customers toward VCF. In September 2026, VMware executives said a new vSphere Standard is coming, with no pricing or date published yet.

Minimum core counts: what applies now

In March 2025, the distributor Arrow told partners that the minimum purchase would rise from 16 to 72 cores per order line from April 10, 2025. Within days there were reports that Broadcom had dropped the change in EMEA, and Broadcom told the press it had “never announced a price change.”

It didn’t stay settled. UK resellers say a 72-core minimum purchase per product has applied to UK orders since December 1, 2025. So the licensing rule is 16 cores per CPU, but the minimum order is a commercial rule that varies by country and channel. Get your reseller to confirm in writing which applies before you model anything.

Renewals, partners and the hyperscaler change

  • Late renewals cost more. Broadcom brought back a 20% surcharge on the first-year price for customers who miss their renewal anniversary.
  • The cloud partner channel shrank. Broadcom ended the white-label model in the EEA, stopped renewing VMware Cloud Service Provider (VCSP) contracts after January 26, 2026, and made the program invite-only. Dropped providers can’t run VMware-based clouds after March 2027. If you use a local VMware hosting provider, check that it’s still authorized.
  • Hyperscalers no longer include VMware licenses. Since November 1, 2025, Broadcom requires a portable VCF subscription on every hyperscaler platform. On Azure VMware Solution (AVS), license-included pay-as-you-go nodes retire on October 31, 2026, and license-included reserved instances on August 30, 2027.

What customers report on price

Published figures vary widely, because the result depends on what you used before, your core density and how you negotiate.

  • CloudBolt commissioned a survey of 302 North American IT decision-makers in January 2026. The median reported increase fell in the 25–49% band, and 86% of respondents said they were actively reducing their VMware footprint.
  • Rijkswaterstaat, the Dutch water agency, estimated that Broadcom’s subscription offer would raise its annual cost by 85%, from €2.14 million to €3.97 million.
  • Tesco says an April 2026 renewal offer was roughly 175% above its 2021 VMware entitlement.
  • CISPE, the association of European cloud providers, cites cumulative increases of more than 1,000% for its members once bundling, up-front payments and minimum commitments are counted.

Europe: complaints, regulators and courts

  • July 2025: CISPE asked the EU General Court to annul the European Commission’s approval of the acquisition.
  • March 2026: CISPE filed a competition complaint with DG Competition, asking for interim measures to suspend the end of the VCSP program.
  • July 2026: CISPE, Beltug, Cigref, VOICE and CIO Platform Nederland jointly asked the Commission for interim measures and a transition period of at least three years. The Commission has sent information requests to cloud providers. As of September 2026 there was no formal case and no interim measures.
  • August 2026: The General Court refused Broadcom’s request to suspend a Commission document demand in the VMware inquiry.

Customers have also gone to court. In June 2025, a Dutch court ordered Broadcom to keep supporting Rijkswaterstaat for up to two years while it moves off VMware. Tesco is suing Broadcom in the UK High Court. AT&T settled its support dispute in November 2024.

The lesson: plan around your renewal date, not around regulatory relief. The EU may eventually change Broadcom’s terms, but it won’t do so on your schedule.

The non-cost reasons to move

Cost starts the conversation. These reasons usually settle it:

  • Support deadlines. vSphere 8 general support ends on October 11, 2027, so staying also means a VCF 9 upgrade project. Three-year contracts signed in 2024 and 2025 expire across 2026 and 2027, close to that date.
  • Hardware refresh. New hosts mean another depreciation cycle, plus VCF cores on every one of them.
  • Datacenter or colocation exit. A facility contract end date is a hard deadline. A cloud move can meet it; swapping the hypervisor can’t.
  • Operating system end of support. SQL Server 2016 reached end of support on July 14, 2026, and Windows Server 2016 follows on January 12, 2027. If you’re rebuilding those servers anyway, rebuilding them on the target platform saves doing the work twice.
  • Skills. Deep VMware skills are getting harder to hire, and one cloud platform team is easier to staff than two stacks.
  • Modernization. Managed databases, containers and AI services are easier to adopt when your data already sits next to them.
  • Resilience obligations. NIS2 and DORA require supplier risk management and exit strategies, whether the supplier is Broadcom or a hyperscaler.

When staying (or moving to another on-premises hypervisor) still makes sense

Lifting workloads to the cloud isn’t automatically cheaper, and it isn’t always the right answer. Keeping at least part of the estate on VMware is defensible when:

  • You run a large, dense, steady-state estate on recent hardware and have a renewal quote you can live with. Steady utilization is where on-premises economics are strongest.
  • Workloads must sit close to something physical: OT systems in plants, latency-sensitive control systems, equipment that can’t move.
  • Regulatory or contractual constraints rule out a hyperscaler for specific data, even after checking EU-region and sovereign options.
  • The renewal is weeks away. A short renewal that buys time for a proper assessment beats a rushed migration.

Moving to another on-premises hypervisor, such as Hyper-V or Azure Local, Nutanix AHV, Proxmox VE or Red Hat OpenShift Virtualization, removes the Broadcom bill. It’s still a migration: you re-tool backup, monitoring, automation and skills, and you keep the hardware and facilities. Tesco’s court filings say its replacement platform doesn’t work with its existing Veeam and Zerto tools. Test the whole toolchain, not just the hypervisor.

Building a TCO comparison you can defend

Most VMware business cases fail the finance review because they compare an old on-premises cost with an optimistic cloud estimate. Compare like for like over the same multi-year period, using your actual renewal quote.

Cost lineWhat to includeCommon mistake
VMware licensingVCF cores at the 16-core floor, extra vSAN capacity, add-ons; portable VCF for any AVS, EVS or VMware Engine hostsLeaving VCF out of the cloud side
HardwareHost and storage refresh, network gear, maintenanceTreating depreciated hardware as free forever
DatacenterPower, cooling, space, colocation contracts, exit penaltiesCounting savings before the facility is fully exited
PeopleOperations staff, third-party support, training, running both environmentsAssuming headcount savings in year one
Cloud compute and storageSizing from measured performance dataSizing from vCenter allocations
Network and egressExpressRoute or Direct Connect circuits, data transfer, firewallsIgnoring internet and cross-region egress
CommitmentsReservations, savings plans, 1- and 3-year host termsCommitting before right-sizing
Microsoft licensingAzure Hybrid Benefit, Software Assurance, Extended Security UpdatesAssuming ESUs are free in Azure for every version
One-off migrationOverlap period of running both, partner services, tooling, testingLeaving out the overlap period

Commitment terms often matter more than list price. Google’s VMware Engine price list shows a ve1-standard-72 node at $9.29 per hour on demand and $3.00 per hour on a three-year upfront commitment, both before the VCF subscription you buy from Broadcom.

Microsoft licensing levers, and one that changed

  • Azure Hybrid Benefit. If you have Windows Server and SQL Server licenses with Software Assurance, Microsoft says you can save up to 85% against pay-as-you-go rates. The benefit includes 180 days of dual-use rights during migration and unlimited virtualization on AVS and Azure Dedicated Host.
  • Extended Security Updates (ESUs). Windows Server 2012/2012 R2 ESUs were free in Azure, including AVS, and delivered their final update on October 13, 2026. Microsoft has since announced that ESU offerings introduced from April 1, 2026 cost the same list price in Azure as elsewhere, and its documentation says SQL Server 2016 ESUs are no longer free on Azure VMs. Windows Server 2016 ESUs bill from January 13, 2027. SQL Server 2014 ESUs stay free in Azure until July 12, 2027.

Don’t put free ESUs for Windows Server 2016 or SQL Server 2016 in a business case without written confirmation from Microsoft.

Azure Migrate and AWS Transform both generate business cases from discovery data. Use them as a starting point and add what they can’t see: people, the facility exit, the overlap period and your real VCF quote.

How: VMware-as-a-service, rehost or refactor

There are three ways to land in public cloud: (a) run VMware on dedicated hyperscaler hosts, (b) rehost VMs to native IaaS, or (c) refactor or replatform applications onto managed services.

VMware-as-a-service options in October 2026

ServiceWho runs the VMware stackVMware licensing nowNotes
Azure VMware SolutionMicrosoft manages the private cloud infrastructure and softwarePortable VCF from Broadcom for all new deployments since November 1, 2025; license-included pay-as-you-go retires October 31, 2026 and reserved instances August 30, 2027HCX Enterprise included; 3 to 16 hosts per cluster; Gen 2 (AV64) hosts attach directly to Azure virtual networks
Google Cloud VMware EngineGoogle manages vCenter, NSX and vSAN lifecyclePortable VCF from Broadcom; required for new commitments bought after October 15, 2025Minimum three nodes for production; 1- and 3-year commitments
Amazon Elastic VMware Service (EVS)You do: install, patch and upgrade VCF yourself or with a partner; AWS runs hosts and networkingBring your own VCF; it can’t be bought through AWSGenerally available since August 2025; VCF 5.2.x, 9.0 and 9.1; i4i.metal and i7i.metal-24xl; Broadcom recommends at least four nodes for production
VMware Cloud on AWSBroadcom, as a managed service on AWS hardwareBought from Broadcom or a Broadcom reseller; AWS stopped reselling on April 30, 2024Still receiving releases (SDDC 1.26 in 2026)

VMware-as-a-service is now a fast way out of your datacenter, but not out of Broadcom. You pay the hyperscaler for whole hosts and Broadcom for VCF cores, and the core count follows the host type: on AVS, each AV64 host counts as 64 cores. Add-ons such as Live Site Recovery and vDefend Firewall also come from Broadcom.

Comparing the three strategies

Criterion(a) VMware-as-a-service(b) Rehost to native IaaS(c) Refactor / replatform
SpeedFastest; VMs move with HCX, no conversionMedium; per-VM replication and cutoverSlowest; done one application at a time
Change to workloadsNone; same VMs, often the same IPsVM conversion, new IPs, driver and OS checksApplication and data changes
Broadcom dependencyRemains (portable VCF)RemovedRemoved
OperationsFamiliar vCenter and NSX toolingCloud-native operations, policy, infrastructure as codeManaged platform services, DevOps
Cost profileWhole hosts plus VCF; storage-heavy estates need extra hostsPer-VM right-sizing, reservations, Azure Hybrid BenefitLowest run cost when done well, highest one-off cost
Best forDatacenter deadlines, hard-to-move legacy, a bridgeMost of the Windows and Linux estateStrategic applications, databases moving to PaaS

A common pattern: VMware-as-a-service as a bridge for what can’t move quickly, native IaaS for most workloads, refactoring for a short list. Give the bridge an end date, or it becomes the new datacenter.

A phased migration approach

1. Discovery and assessment

On Azure, the Azure Migrate appliance is a VM in vCenter that collects configuration and performance data continuously. One appliance covers up to 10,000 VMs across up to 10 vCenter Servers. For air-gapped networks, Azure Migrate Collector takes a one-off snapshot. An RVTools XLSX import (preview) gives a quick business case without an appliance, but without performance data. RVTools was hit by a supply-chain attack on May 12, 2025, so download it only from the official site and check the hash.

On AWS, AWS Transform runs an agent-based workflow, now called “Migrations (including VMware)”, fed by Application Discovery Service collectors, the Export for vCenter tool or imported files such as RVTools. Since May 2026 it also produces a TCO and business case. Migration Evaluator and the no-cost Optimization and Licensing Assessment (OLA) are still offered.

Whichever tool you use, collect performance data across a full business cycle, including month-end and quarter-end peaks, before you size anything.

2. Dependency mapping

Azure Migrate offers agentless and agent-based dependency analysis, and AWS Transform builds dependency graphs. Both see network connections, not business logic. Add firewall logs, load balancer configurations, the CMDB and application owner interviews, and look for monthly jobs a short observation window will miss. The lessons learned from application migrations post covers this in detail.

3. Business case and a decision per workload

Classify every workload as retire, retain, VMware-as-a-service, rehost or refactor, and have finance sign off the assumptions. Size the next VCF renewal for what will still be on VMware at the end of the term, not what runs today.

4. Landing zone readiness

Identity, connectivity, policy, logging, security tooling and backup must be in place before wave one. AVS also needs an EA or MCA subscription, approved host quota (up to five business days) and a non-overlapping /22 management block.

For GDPR and data residency, confirm the region for the workloads and also where backups, logs and replicas end up. The datacenter migration to Azure post covers landing zone design.

5. Wave planning

Group servers that talk heavily to each other and can share a change window. Start with low-risk, non-production workloads to prove the runbook, and don’t leave chatty tiers split across the interconnect longer than necessary. Plan around change freezes and the business calendar.

6. Migration tooling

TargetToolNotes
Azure VMware SolutionVMware HCXHCX Enterprise at no additional cost; bulk, vMotion, Replication Assisted vMotion, cold and OS-assisted migration
Amazon EVSVMware HCXPrivate connectivity through Direct Connect or Site-to-Site VPN via Transit Gateway, or over the public internet with Elastic IPs
Azure VMsAzure Migrate (agentless or agent-based)Agentless: up to 300 VMs replicating at once per vCenter with one appliance, 500 with a scale-out appliance
Amazon EC2AWS Transform MGNRenamed from AWS Application Migration Service in June 2026; continuous replication; free for 90 days per server

Snapshot-based backups conflict with Azure Migrate’s agentless replication, so plan backup windows and snapshot clean-up first.

7. Testing and cutover

Run test migrations into an isolated network and have application owners sign them off. Use a cutover runbook with go/no-go and rollback criteria, lower DNS TTLs in advance, swap monitoring and backup agents, and keep application teams on call during hypercare.

8. Decommission

Keep source VMs powered off but recoverable until hypercare ends, then delete them and sanitize the storage. Cancel hardware support and reduce VCF cores at renewal. Savings only appear when you stop paying.

Networking, storage, backup and DR

Networking

HCX network extension lets VMs keep their IP addresses while moving to AVS or EVS. Treat it as temporary and remove it once a segment’s last VM has moved. AVS Gen 1 reaches on-premises through ExpressRoute Global Reach, while Gen 2 uses a standard ExpressRoute connection or Site-to-Site VPN. EVS needs Transit Gateway in front of Direct Connect or VPN.

Native IaaS doesn’t stretch your VLANs, so most VMs get new IP addresses. Find hard-coded IPs in configs, firewall rules and IP-bound licenses early, and clean up DNS first. AWS Transform can translate Cisco ACI, Palo Alto and FortiGate configurations into a VPC design.

Storage

On VMware-as-a-service, vSAN capacity often decides the host count, not CPU. External datastores separate storage from compute: Azure NetApp Files for AVS, FSx for NetApp ONTAP for EVS. On native IaaS, size disks for IOPS and throughput, not just capacity.

Backup and DR redesign

VMware-integrated backup and DR products run on AVS through partners, who also support them; Live Site Recovery needs a Broadcom license. On native IaaS, redesign around Azure Backup and Azure Site Recovery, or the AWS equivalents, using availability zones and a second region. Then test restores and failover, because DORA expects tested resilience.

Exit planning also gets easier: from January 12, 2027, the EU Data Act bans switching charges between cloud providers.

Operating model changes

  • Who does what. On AVS and Google Cloud VMware Engine, the provider manages the VMware software and you configure NSX, HCX and your VMs. On EVS, you own the VCF lifecycle. On native IaaS, a platform team owns the landing zone.
  • License compliance. On AVS, registered VCF cores must exactly match deployed BYOL cores. Running more than Broadcom licensed puts the private cloud at risk of suspension.
  • FinOps. Give every workload a cost owner, enforce tagging and actively manage reservations.
  • Skills. Infrastructure as code and policy replace click-ops in vCenter. Retrain your VMware administrators: they know the estate and make the best migration engineers.

Funding programs

Microsoft has relaunched Azure Accelerate as Frontier Accelerate for Azure for its FY27, with Cloud Accelerate Factory delivery help (VMware-to-AVS migration is listed), Azure credits and partner funding, subject to eligibility. The AVS reserved-instance credit under the VMware Rapid Migration Plan ended on June 30, 2026.

AWS offers the Migration Acceleration Program (MAP) for VMware, with credits or partner investment for qualified migrations, plus the no-cost OLA. With either provider, raise funding before you sign anything.

Common pitfalls

  • Treating AVS, EVS or Google Cloud VMware Engine as an exit from Broadcom.
  • Sizing from allocated vCPU and RAM instead of measured performance.
  • Under-scoping dependency mapping and discovering a monthly interface after cutover.
  • Assuming free ESUs for Windows Server 2016 and SQL Server 2016.
  • Leaving HCX network extensions in place permanently.
  • Signing a multi-year VCF renewal for the whole estate when you plan to shrink it.
  • Committing to three-year reservations before right-sizing.
  • Underestimating lead times for AVS host quota, ExpressRoute or Direct Connect circuits, and Broadcom license procurement.
  • Migrating without a platform team and operating model ready for day two.

Where to start

  1. Put every deadline on one page: VMware renewal date, vSphere 8 end of support (October 11, 2027), hardware end of life, facility contracts, Windows Server 2016 (January 12, 2027), and AVS license-included retirement if it applies to you.
  2. Get written one-year and multi-year renewal quotes, with core counts and the minimums that apply in your country.
  3. Run discovery with performance data. Start with an RVTools import for a first view, then deploy the Azure Migrate appliance or AWS Transform discovery.
  4. Build the TCO with every cost line above and get finance to sign off the assumptions.
  5. Classify each workload: retire, retain, VMware-as-a-service, rehost or refactor.
  6. Check landing zone readiness against your NIS2, DORA and GDPR obligations.
  7. Talk to Microsoft or AWS about funding before you sign contracts.
  8. Run a pilot wave of low-risk, non-production VMs, then size the VCF renewal to what will remain.

Sources

Last reviewed: October 2026.

Leave a Reply

Rares Juretchi

Rares Juretchi

Rares is a cloud and enterprise architect with over 25 years of IT experience. He has led or architected over 100 cloud programs in regulated industries, including healthcare, finance, manufacturing and energy, and specializes in Azure landing zones, migrations and governance, with AWS alongside.

The blog offers practical guidance on:

  • Cloud migration strategy and execution
  • EU digital regulation (NIS2, DORA, the AI Act and more) for cloud teams
  • Cloud security frameworks and compliance
  • Landing zones and cloud governance

Read the full profile →

Let’s connect