The best hybrid cloud platform for running legacy VM workloads on-premises and in the cloud

The problem: legacy VMs, a licensing shock and two places to run them

Most large organisations still run thousands of virtual machines that nobody plans to rewrite: packaged ERP, line-of-business Windows applications, databases tied to specific operating system versions, and vendor appliances. For a decade the default home for these workloads was VMware vSphere, and the default hybrid story was to run the same vSphere stack in a public cloud.

That default has been shaken. Since Broadcom acquired VMware, licensing has moved to per-core subscriptions bundled into VMware Cloud Foundation (VCF), and the hyperscalers have stopped bundling VMware licences into their own VMware services. Australian share registry Computershare told The Register it was quoted a renewal that represented an increase by a factor of between 10 and 15, and that it would move 24,000 VMs to Nutanix. Speaking at Gartner's Symposium in Australia, Gartner analyst Julia Palmer predicted that more than a third of workloads running under VMware will run on another platform by 2028.

The technical problem underneath is harder than switching hypervisors. A legacy estate needs a platform that runs old guest operating systems reliably, converts VMs without long outages, and gives operations teams one control plane across the data centre and at least one public cloud. It also needs a licensing model that finance can forecast and that does not simply recreate the lock-in the organisation is trying to escape.

Who this is for

• Heads of infrastructure and enterprise architects with hundreds or thousands of on-premises VMs, usually on vSphere, facing a VMware renewal decision.
• Organisations that want to keep legacy workloads on-premises for latency, data residency or cost reasons, but also run some of the same workloads in AWS or Azure.
• Australian and New Zealand buyers who need the cloud side of the platform hosted in an Australian region.
• It is not aimed at teams already refactoring applications into containers; for them, the Kubernetes platform matters more than the VM platform.

How we evaluated

We set the criteria and weights before research, based on the problem above. We then reviewed official documentation, pricing pages, regulatory filings and independent reporting for seven platforms, recording a source for every finding. We re-checked the decisive facts (migration tooling, pricing models, cloud regions and hardware rules) against the live pages on 20 October 2026. Each platform was scored from 1 to 5 on each criterion with a written evidence note.

What we did not do: we did not run these platforms, migrate test workloads, or request quotes. Prices are public list prices in US dollars where a vendor publishes them; most vendors here do not, and we say so. Where a vendor makes a claim we could not confirm independently, we attribute it.

Side-by-side comparison

Why Nutanix Cloud Platform is the champion

It is the same platform in your data centre and in the cloud. Nutanix Cloud Clusters (NC2) runs the Nutanix stack on bare-metal servers in public clouds. Microsoft's documentation for NC2 on Azure states that clusters run only the AHV hypervisor, with Prism Central deployed to manage them, in clusters of three to 28 nodes. NC2 is also available on AWS. Nutanix Central is described by Nutanix as a unified console for Nutanix services on-premises or on NC2, with role-based access across data centres and public cloud bare metal. Unlike the hyperscaler VMware services, the customer, not the cloud provider, controls the software lifecycle.

It has a mature path off vSphere. Nutanix Move migrates VMs from ESXi to AHV. The Move 6.3.3 release notes, posted on 13 October 2026, list ESXi 6.5 through 8.0 Update 3 and vCenter up to 9.0 as sources, and add a DirectPath mode that transfers disk data directly from ESXi to AHV. This matters for legacy estates, which often still run older ESXi hosts. Independent evidence supports the path at scale: Computershare's 24,000-VM move, and Gartner's advice, reported by The Register, that VMware users should consider Nutanix first.

Licences move with the workload. Nutanix says on its NC2 FAQ that licences are portable across clouds, so no new licences are needed when migrating VMs, and offers pay-as-you-go billing per core-hour and SSD-hour as well as committed terms. Its FY2026 annual report says licences typically run from one to five years and that the software runs on qualified hardware from Cisco, Dell, Fujitsu, HPE and Lenovo, so most organisations can reuse server relationships they already have.

It can keep the cloud side in Australia. NC2 on Azure is listed in Australia East, and Nutanix lists Asia Pacific (Sydney) among NC2 on AWS regions. Nutanix is also financially solid for a long platform commitment: its FY2026 results report revenue of US$2.85 billion, up 12%.

Nutanix did not win every criterion. VMware Cloud Foundation is more compatible, because it needs no conversion at all, and Azure Local is more transparent on price. Nutanix won because it scored 4 or 5 on every criterion except pricing, and because it is the only alternative to VMware that offers the same stack on-premises and in two major clouds.

Where Microsoft Azure Local is the better choice

Azure Local, formerly Azure Stack HCI, finished second and is the better pick for organisations whose cloud strategy is already centred on Azure. Its VMware migration path is built into Azure Migrate: Microsoft says the VMware to Azure Local migration requires no preparation of source VMs, including no agents, keeps the data flow local, and is run from the Azure portal. Once migrated, VMs are managed through the Arc resource bridge with Azure role-based access control, tags and the same Azure Resource Manager control plane used for cloud resources.

It is also the most transparent on cost. Microsoft's billing documentation describes a flat rate per physical core with no minimum and no maximum, billed on the Azure invoice, so existing Enterprise Agreement discounts and Azure commitments apply. Microsoft's public Azure Retail Prices API lists the standard service fee at US$0.33 per core per day, about US$10 per core per month. The pricing page lists a Windows Server guest subscription at US$23.30 per physical core per month, a 60-day free trial, and an Azure Hybrid Benefit that waives both fees for eligible Windows Server Datacenter customers on the standard tier.

Choose Azure Local if most of your legacy estate is Windows Server, you already hold Windows Server Datacenter licences with Software Assurance, and your public cloud is Azure. Expect trade-offs: Microsoft supports only hardware listed in the Azure Local catalogue; the standard hyperconverged tier supports up to 16 nodes, which Gartner described as well short of the fleets many VMware users run; Windows Server 2012 and 2012 R2 guests do not support guest management; and the cloud side is Azure's own VM service, not Azure Local, so consistency comes from the control plane rather than the same stack.

Other credible options

VMware Cloud Foundation (3.65) remains the most compatible choice: existing VMs run unchanged on-premises and in Azure VMware Solution, Google Cloud VMware Engine and Amazon Elastic VMware Service, all with Australian regions. It scored lowest of all on cost predictability. Broadcom publishes no list prices; its core-counting rules require at least 16 cores per CPU; The Register reported a 72-core minimum order from April 2025 that VMware did not confirm. Microsoft no longer includes a VCF licence with new Azure VMware Solution nodes, requires existing pay-as-you-go customers to bring their own by 31 October 2026, and says it reports those registrations to Broadcom monthly. Stay on VCF if the renewal quote is acceptable and compatibility outweighs everything else.

Red Hat OpenShift Virtualization (3.30) is the best option for organisations that want legacy VMs and containers on one Kubernetes platform. Red Hat certifies Windows Server 2016 to 2025 and RHEL 7 to 10 guests, its migration toolkit supports warm migration from vSphere, and the same platform runs on ROSA, ARO and Oracle Cloud according to its installation documentation. It lost ground on pricing, which is not published, on bare-metal requirements, and on Windows Server 2012, which is not certified.

HPE Morpheus VM Essentials (3.15) is the budget challenger: a KVM hypervisor licensed per socket rather than per core, with a recommended price reported at US$600 per socket per year. We could not reach HPE's own pages to confirm the price or the official guest operating system list, and its hybrid story is a management overlay rather than the same stack in the cloud.

AWS Outposts (3.00) brings the same AWS APIs and tools into your data centre and is supported in Australia and New Zealand with Sydney as the parent Region. It is a poor fit for legacy VMs: every VM must be rebuilt as an EC2 instance, AWS's migration service requires Windows Server 2008 R2 or later on Nitro, hardware comes only from AWS, and AWS says racks are not designed for disconnected operations.

Google Distributed Cloud (2.30) is the only option here with a fully air-gapped offering, which matters for defence and some government workloads. For general legacy estates it is weak: VMs are managed with Kubernetes tools, we found no documented vSphere migration tool, the connected edition requires a 36 or 60 month commitment with hardware that cannot be resized, and it is not offered in New Zealand.

Limitations and trade-offs

• Nutanix still means a hypervisor change. Every VM moves from ESXi to AHV, which changes drivers and boot behaviour. The Move release notes record a known issue where some Secure Boot Linux guests fail to boot after migration. We could not verify the current AHV guest operating system list for very old systems such as Windows Server 2008; check it on the Nutanix support portal before planning.
• Nutanix does not publish prices. Its portable licences and pay-as-you-go option are flexible, but buyers must model costs from quotes. Get list pricing and renewal caps in writing.
• Azure Local's published fee is not fully visible. On 20 October 2026, Microsoft's pricing page did not render the amounts for the standard and advanced tiers; the US$0.33 per core per day figure comes from Microsoft's retail price API under the older Azure Stack HCI name. Billing also continues for up to 31 days after a cluster is shut down unless its Azure resource is deleted.
• New Zealand hosting is thin across the board. Neither NC2 on Azure nor Azure VMware Solution lists New Zealand North; NC2 on AWS in New Zealand was in private preview in October 2025, according to Reseller News. New Zealand buyers needing local cloud hosting should confirm current status.
• Scores reflect documentation, not performance. We did not measure migration speed, VM density or failure behaviour. A pilot on representative workloads is essential.

Implementation and pricing considerations

1. Inventory before you choose. Export a list of every VM with guest operating system version, vCPU, memory, disk and dependencies. The oldest operating systems and the largest VMs decide which platforms are viable.
2. Model costs on your real core and socket counts. As a worked example on Azure Local list prices, a four-node cluster with two 16-core processors per node has 128 physical cores: about US$1,280 a month in host fees and US$2,982 a month if you also take Microsoft's Windows Server guest subscription, before hardware and support. Under the reported HPE price, the same eight sockets would cost about US$4,800 a year. Nutanix and Broadcom require quotes.
3. Run the migration tool against your oldest VMs first. Test Nutanix Move, Azure Migrate or the OpenShift migration toolkit on a sample that includes Secure Boot guests, old Windows Server releases and multi-disk databases.
4. Decide where the cloud side lives. If you need a public cloud copy of the platform in Australia, confirm the region and instance type is available for your size: NC2 on Azure lists a single node type in Australia East.
5. Plan for licence portability and exit. Ask how licences transfer between on-premises and cloud, what happens at term end, and how VMs can be exported in a standard format if you leave.

Questions to ask vendors

1. Which guest operating system versions are supported today, including Windows Server 2012 R2 and older Linux releases, and what is the support policy when they reach end of life?
2. What downtime should we expect per VM when migrating from vSphere, and which VM configurations does your migration tool not support?
3. Can the same licence be used on-premises and in a public cloud, and what does moving it cost?
4. What are your list prices per core or per socket, the minimum purchase, and the maximum renewal increase you will commit to in writing?
5. Which management, security and backup features differ between your on-premises platform and the cloud version?
6. Which Australian and New Zealand regions offer the cloud version today, and with which node types?
7. Can the platform run disconnected from your cloud for extended periods, and what stops working if it does?
8. How do we export VMs in a standard format if we leave your platform?

What would change our verdict

We will revisit this recommendation if Nutanix publishes list prices or changes its licence portability terms, if Microsoft raises the 16-node limit on its standard Azure Local tier or offers Azure Local as a hosted service in Azure, if Broadcom publishes VCF prices or restores licence-included hyperscaler options, or if NC2 or another platform reaches general availability in a New Zealand region. Each change would move at least one score on the criteria above.