AOS 7.6 and Prism Central 7.6 both landed on 27 July 2026. Skim the highlights and it looks like a capacity bump with some dashboard work attached. Spend longer with it and it’s more than that — the density numbers change how you size a cluster, the external storage work changes who Nutanix competes with, and there’s a deprecation list that will quietly break automation in a lot of environments.
This is the architectural read rather than the feature list.
Density has outgrown the old failure-domain assumptions
The headline numbers: 307TB per node on all-flash NVMe, up to 40 EDSFF NVMe drives in a single node, and 270TB on hybrid.
Capacity per node has been climbing for years, but this is the point where the old sizing instincts stop working. A cluster that used to need eight nodes for capacity might now need three or four. That’s excellent for rack space, power and licensing — and it quietly makes every remaining node a much bigger blast radius. Lose one node out of twelve and you shrug. Lose one out of four, when that node held 300TB, and the rebuild is a different conversation entirely.
Nutanix clearly saw the same problem, because the fault tolerance work in this release is the direct answer:
- 1 Node / 2 Disk (1N/2D) is a new fault tolerance option — tolerating a node loss plus two disk failures. With 40 drives in a chassis, the odds of a disk failing while you’re already a node down stop being theoretical.
- 1 Node / 1 Disk (1N/1D) has been extended to larger clusters.
My take: don’t size on the new maximum just because you can. Work out what a rebuild of one fully-populated node actually costs you in time and front-end performance, and let that set your node count. Density is a budget you spend on resilience, not only on rack units.
Two-node clusters finally have a way forward
Small but genuinely useful: two-node AHV clusters can now expand to three nodes, with a simplified workflow in Prism Central.
Anyone who has deployed two-node clusters for branch or remote sites knows the trap. They’re the right answer at the time, then the site grows, and adding capacity meant a rebuild. Making that an expansion rather than a migration changes whether two-node is a sensible starting point or a corner you paint yourself into. For distributed estates — retail, branch banking, industrial sites — this one matters more than its size in the release notes suggests.
External storage is becoming a strategy, not an exception
7.5 brought Pure Storage FlashArray. 7.6 adds Dell PowerStore over NVMe/TCP, managed directly from Prism Central, with disaster recovery alongside it — Metro Availability, synchronous and asynchronous replication, and live VM migration.
Two releases in a row is a pattern rather than a one-off. Nutanix is making it viable to run its compute and management layer over storage it didn’t sell you, and it changes the shape of a lot of migration conversations. The usual blocker is a customer three years into a five-year array investment. Being able to bring in Nutanix now and revisit the storage later removes the reason to say “not yet” — and the DR feature set means it isn’t a compromise deployment.
Prism Central did the unglamorous work
Most of the operational improvements in this release live in PC 7.6, and several of them remove daily friction:
- Single Cluster Profile for managing storage configuration across multiple clusters, and reusable VM Profiles for standardised deployments — configuration as a template rather than a habit.
- Paste into the VM console, passwords included. A small thing that anyone who has hand-typed a complex password through a console will appreciate.
- Resumable image and OVA uploads with progress tracking, which makes large uploads over unreliable links survivable.
- Centralised storage dashboard for health, capacity and performance, plus a dedicated management page for Controller VMs.
- Higher concurrent user session limits, scaled to deployment size.
On the DR side: automated recovery plans for Zero Compute DR in AWS and Azure, a rebuilt DR dashboard, more flexible IP and DNS mapping in recovery plans, age-based bulk deletion and automatic expiry of obsolete recovery points, and higher limits for protected VMs, volume groups and recovery points.
The one I’d single out is the guided migration tool from legacy Protection Domains. Plenty of environments still run PD-based protection because migrating off it was a manual project nobody scheduled. A guided path removes that excuse, and it’s worth planning in.
Security and multi-tenancy
Hardening gets easier. Predefined Security Profiles let you apply a cluster hardening posture as a single selection rather than assembling it setting by setting. Secure Access brings SSH access control including support-restricted login — which answers the “who can reach the shell, and when” question that comes up in every audit. Encryption keys get automatic external key health checks and regeneration, and the Security Dashboard now extends to NC2 on both AWS and Azure.
There’s also a new Nutanix Projects framework for organising workloads, with project-specific categories that no longer collide on naming, and automatic workload migration during upgrade. If you’ve ever tried to keep a category taxonomy coherent across teams, the naming-conflict fix alone is worth having.
The part everyone skips: deprecations
This is where 7.6 will actually hurt people, so read it before you plan the upgrade.
Legacy Nutanix APIs v0.8 through v3 are deprecated, with v4 as the migration target. If you have automation — Ansible playbooks, backup integrations, custom reporting, that one PowerShell script running on a schedule since 2021 — a good portion of it is likely still on v2 or v3. Deprecated isn’t removed, but this is the notice period, and the time to audit is before you’re forced into it.
Also going: the Console tab, replaced by Launch Console, and cluster-level quotas for Projects and Self-Service are flagged for future removal.
My advice: grep your automation repos for API version strings this month, not the week before you upgrade. That’s a small task now and a bad surprise later.
Platform housekeeping
Controller VMs and Prism Central move to a Red Hat Enterprise Linux 9 base — a foundation change with a long tail of security and support benefit. Live migration from older AHV versions runs with zero downtime, which eases mixed-version transitions. VMware vCenter Server 9.0 with ESXi 8.0 is supported for anyone still running mixed hypervisor estates. And the interface now aligns to WCAG 2.2 AA accessibility standards, which is a procurement requirement in public sector work more often than people expect.
What I’d do with this
Three things, in order. Audit your automation for deprecated API versions — that’s the item with a deadline attached. Re-run your node sizing against the new density figures and the 1N/2D option, rather than carrying forward last year’s assumptions. And if you have two-node sites you’ve been treating as fixed, revisit whether that’s still true.
If you’re working out which release to target and when your current build leaves support, I wrote about that in more detail here. And if you’d like a second opinion on sizing or the upgrade path for your own environment, get in touch.