Cisco HyperFlex Is Being Wound Down: the Nutanix Path, and the Costs Nobody Quotes

Written and maintained by Haink's infrastructure team · Verified against Cisco bulletins, ordering guides and the migration white paper, 22 August 2026 · authorized-channel, serial-verified

HyperFlex did not get an end-of-life announcement. It got several — one for the software, one for each hardware generation, one for components — and the effect of them together is that the platform is closed. Cisco's own software advisory puts it about as plainly as a vendor ever does: customers should "plan and start their migration from HyperFlex to Nutanix HCI on Cisco UCS, or another suitable solution, as soon as possible." Under the heading of workaround, the advisory says: "None."

The destination Cisco points at is real and well documented. What is not well documented — and what determines the budget — is that the migration costs more than the hardware line suggests, for three specific reasons covered below.

Where things actually stand

Software maintenance for HXDP ended on 11 September 2025; after that the advisory describes best-effort TAC only, through to February 2029. The releases affected span HXDP 2.0(x) through 6.0(x) — that is, all of them.

One inconsistency to be aware of, because it comes up in vendor conversations: Cisco's HyperFlex end-of-life FAQ contains a line stating that HyperFlex M6 and UCS M6 "have not reached end of sale or end-of-life." That contradicts the EOL15171 and EOL15369 bulletins, which announce exactly that for the M6 nodes. Where a FAQ and a numbered bulletin disagree, the bulletin is the formal document.

What Cisco offers instead

The named destination is Cisco Compute Hyperconverged with Nutanix — Cisco UCS servers running the Nutanix Cloud Platform. Each node carries three software layers: UCS firmware, the Nutanix AHV hypervisor, and Nutanix AOS storage. Hardware and firmware are managed through Cisco Intersight; storage and hypervisor through Nutanix Prism.

The nodes are distinct part numbers rather than ordinary UCS servers — HCIAF220C and HCIAF240C on M6 and M7, HCINX220C and HCINX240C on M8, in rack and X-Series forms. Minimum cluster size is three nodes, with one- and two-node options for edge and branch sites.

Support is cooperative and, unusually, actually specified: if you bought hardware and software through Cisco, Cisco is the first point of contact. Cisco verifies entitlement for hardware cases, Nutanix for software cases, and the two ticketing systems are eBonded so a case is visible to both. Cisco TAC owns hardware, network and Intersight; Nutanix owns software, installation, upgrades and licensing.

Cisco names two other options in the EOL FAQ, without elaborating: repurpose the HyperFlex hardware as standard Cisco UCS servers and leave hyperconverged behind, or run other supported HCI software on it. Worth knowing that Cisco also sells Cisco Compute Hyperconverged with VMware vSAN on the same server family — though Cisco does not name it as a HyperFlex migration path in any end-of-life document.

The three costs that are not in the hardware quote

1. The Nutanix licence is a new purchase, with no credit

This is unambiguous in Cisco's FAQ. Migration happens "with the purchase of a new Nutanix software license," and on the question of credit for what you already own: "While there will be no direct credit for existing HyperFlex licenses, Cisco and Nutanix are working on a number of migration offers to ease the cost of switching to the new platforms."

Nor is there a refund for unused HyperFlex subscription term — those are right-to-use subscriptions that "can be used until the subscription expires or last date of support, whichever comes first."

Two further points that shape the number:

• Nutanix licensing is per processor core. The licence quantity must match the core count ordered, so a core-dense refresh raises the software line proportionally.
• Compute-only nodes still require a licence. A common assumption in sizing is that only storage nodes are licensed. They are not.
• Replication, metro clustering and data-at-rest encryption require NCI Ultimate, or Pro with an add-on. If the HyperFlex estate had DR, that capability is a separate budget line rather than something included.

On offers: Cisco's own "Refuse to Lose" programme, which gave free M6 and M7 nodes with special -FRE part numbers, was a fiscal-2024 promotion and has expired. The current Nutanix migration promotion targets VMware, Oracle and OpenShift customers and does not name HyperFlex. As of August 2026 there is no published HyperFlex-specific offer — Cisco directs customers to their account team, which means the discount is negotiated rather than listed.

2. Reusing your existing hardware means moving the data twice

Cisco says qualified HyperFlex M6 nodes can be repurposed. That sounds like the cheap path, and it is — until you read how it works. From the EOL FAQ, the procedure on existing hardware is to "evacuate all virtual machines and data from the HyperFlex cluster to another storage platform. Build the new Nutanix cluster on the existing HyperFlex server hardware and migrate the data and virtual machines back."

There is no in-place conversion. The workload leaves, the cluster is rebuilt, the workload comes back. That means an interim platform sized for the entire estate, two full migration passes, and a maintenance window proportional to the data volume. On a large cluster the interim storage alone can cost more than the difference between repurposing and buying new nodes — which is why the "free hardware" path frequently turns out not to be the cheaper one once it is properly costed.

Two hard constraints on top:

• M5 cannot be repurposed. Cisco states that the HyperFlex M5 platform carrying an HX hardware PID "is currently not supported with Cisco Compute Hyperconverged with Nutanix." For M5 estates the choice is new nodes, or repurposing the servers as standard UCS and leaving HCI.
• Which M6 nodes qualify is defined by Nutanix, not Cisco. The FAQ does not list part numbers; it refers you to the Nutanix hardware compatibility matrix. Check the specific PIDs before assuming a node qualifies. Repurposing also requires clearing the HyperFlex personality from the server through CIMC.

3. Snapshots have to go before anything moves

A detail that has stopped more than one migration window: every HyperFlex native snapshot must be deleted before a VM can be moved. VMware cannot vMotion a virtual machine with storage-backed native snapshots attached, so the deletion is a prerequisite, not a cleanup step.

And if you migrate with Nutanix Move — the free tool, which has the advantage of leaving the source VM intact — all existing snapshots of the migrated VM are deleted in the process. Anything depending on those snapshots for recovery points needs a different plan before the migration starts, not after.

Hypervisor: what happens to your VMware licences

AHV is included with the node — it is one of the three software layers and carries no separate licence. VMware ESXi is also supported on these nodes, but the licence is your own and it comes from Broadcom: Cisco's guidance is that existing ESXi licences can run until expiry, after which renewal is a Broadcom conversation.

For estates already re-evaluating VMware costs, this is the moment that decision collides with the HyperFlex one. Moving to AHV removes a licence line entirely; staying on ESXi keeps it and adds the Nutanix one. It is worth pricing both before the hardware is ordered, because the answer changes what you buy.

How to plan this

1. Split the estate by generation. M5 has no repurposing path; M6 may. That single fact reorganises the project.
2. Check the specific M6 part numbers against the Nutanix compatibility matrix before assuming reuse is available.
3. Cost the interim storage honestly if repurposing. Two migration passes and a temporary platform frequently exceed the price of new nodes.
4. Price the Nutanix licence by core count, including compute-only nodes, and check whether DR features push you to Ultimate.
5. Settle the hypervisor question — AHV or ESXi — before ordering, because it changes both the licence stack and the migration tooling.
6. Audit snapshots and DR dependencies before scheduling any window.
7. Ask the account team about migration offers. Nothing is published, which means what is available depends on asking.

Frequently asked questions

Is Cisco HyperFlex end-of-life?

Yes, through a series of bulletins rather than one. The HXDP software went end-of-sale in September 2024, M6 nodes in March 2024 and M5 nodes in October 2023. Support runs to February 2029 for the software and to 2028–2031 for the hardware depending on generation.

Can I reuse my HyperFlex servers for Nutanix?

Qualified M6 nodes, yes — check the specific part numbers against the Nutanix hardware compatibility matrix, and expect to clear the HyperFlex personality from the server first. M5 nodes carrying an HX hardware PID are not supported.

Does my HyperFlex licence carry over to Nutanix?

No. A new Nutanix software licence is required, and Cisco states there is no direct credit for existing HyperFlex licences. There is also no refund for unused subscription term.

Can the data stay in place while the cluster converts?

No. Cisco's documented procedure is to evacuate all virtual machines and data to another storage platform, rebuild the cluster on the existing hardware, then migrate back. Plan for interim storage and two migration passes.

Do I have to move to AHV, or can I keep VMware?

Both are supported. AHV is included with the node; ESXi requires your own licence, renewed through Broadcom once the current one expires. Since VMware licensing is under review at most organisations anyway, it is worth pricing both paths before ordering.

What if I do not want hyperconverged infrastructure at all?

Cisco explicitly allows repurposing HyperFlex hardware as standard UCS servers. For estates where the HCI model was never the point — where it was bought as a convenient way to get compute and storage in one purchase — that is often the cleanest exit.

Related

• Cisco end-of-life guide — every family in migration, with the binding dates
• Nutanix vs VxRail — the target architecture compared
• Hyperconverged infrastructure — when HCI is and is not the right model
• How to choose Cisco switches, routers and firewalls · NVIDIA · Dell
• Enterprise servers · Storage systems · Cisco stock and lead times
• Sovereign cloud infrastructure — 40× AMD EPYC 9654, 7,680 vCPUs, 96 TB NVMe

Sources

• Cisco — HyperFlex software advisory ("migrate… as soon as possible", workaround "None")
• Cisco — HyperFlex end-of-life FAQ (new licence required, no credit, migration procedure, M5 not supported)
• Cisco — HXDP software end-of-life bulletin · M5 nodes bulletin
• Cisco — Migrating workloads from HyperFlex to Compute Hyperconverged with Nutanix (migration tooling, snapshot constraints)
• Cisco — Cisco and Nutanix cooperative support
• Cisco — Compute Hyperconverged with Nutanix ordering guide (M8) (per-core licensing, node PIDs)
• Cisco UCS hardware compatibility tool