Summary: The date was April 7, 2026, the stage was .NEXT 2026 in Chicago, and Nutanix wagered its full product line on one claim: enterprises should be able to run production-grade agentic AI inside their own data centers. Cloud Manager 2.0 puts one control plane across sites and domains — air-gapped and heavily regulated environments included — with cost governance, AIOps, and self-service built in. GPU supply remains tight, so steering customers toward hardware they already run is an explicit part of the sales case.
The date was April 7, 2026, the stage was .NEXT 2026 in Chicago, and Nutanix wagered its full product line on one claim: enterprises should be able to run production-grade agentic AI inside their own data centers. The latest Nutanix Cloud Platform (NCP) capabilities pull compute, storage, networking, and AI services into a single stack. SiliconANGLE framed it plainly — a vendor that grew up in hyperconverged infrastructure is recasting itself as an agent platform company for hybrid and multicloud settings.
For teams bleeding budget on cloud AI and losing sleep over data sovereignty, the instruction fit on one line: move the agents to where the data lives, not the data to where the agents run.
Nutanix Agentic AI: Tenant Isolation with a Watch on Token Bills
Nutanix Agentic AI, the flagship, is slated for the second half of 2026 and arrives alongside a fresh multitenant AI management layer built for “neoclouds” — the emerging breed of providers selling AI infrastructure on demand. That layer lets a provider climb from GPU-as-a-service into Kubernetes-, model-, and notebook-as-a-service; tenants on shared GPU clusters stay separated from one another, and LLM token consumption is kept under watch so the month-end invoice holds no surprises.
The supporting lineup runs deep. Service Provider Central tightens tenant isolation, resource management, and service catalogs. Cloud Manager 2.0 puts one control plane across sites and domains — air-gapped and heavily regulated environments included — with cost governance, AIOps, and self-service built in. NC2 stretches into more hyperscaler regions and sovereign environments, among them the AWS European Sovereign Cloud and the coming Google Cloud C3 bare-metal instances with Hyperdisk.
NKP Metal: Kubernetes Straight on Bare Metal, and the “Dual-Native” Claim
NKP Metal skips the hypervisor and drops Kubernetes onto bare metal, aimed at AI training, edge deployments, and dense GPU configurations. On training clusters, hypervisor overhead is pure waste — so this is the line between renting extra GPUs and actually putting the ones you already own to work. Foundation and Lifecycle Manager handle the automation, and Cloud Native AOS data services now reach down to physical infrastructure. Lee Caswell, SVP of product and solutions marketing, dubs the design a “dual-native” architecture — containers can sit on VMs or on bare metal, all governed by one management and security framework — and asserts Nutanix alone offers the approach. In his words: “We’re bringing the simplicity that Nutanix has been known for into bare metal.”
The conference theme kept the same economy: “one platform and one experience only.”
Migration and Ecosystem: Zero-Copy Moves, a Bench of 100-Plus Partners
For companies plotting an exit from VMware, Nutanix put zero-copy migration on the table: vSphere workloads turn into AHV virtual disks in place, and no data gets copied along the way. The argument has teeth because migration cost is the standard excuse for standing still — converting in place spares a large estate the storage-duplication tax that quietly doubles the project budget. On storage, NetApp ONTAP external storage support lands in the second half of 2026, Dell PowerStore integration is penciled onto the roadmap, and NetApp has struck a strategic alliance meant to carry ONTAP into the Agentic AI stack.
The ecosystem ledger is sizeable. .NEXT 2026 drew more than 100 partners under one roof for the first time; Dell walked away as 2026 Global OEM Partner of the Year, and Palo Alto Networks as Global Security Partner of the Year. The register also lists Cisco’s AI PODs validated designs, AHV on Lenovo ThinkSystem, NVIDIA AI Enterprise, a multi-year AI partnership with AMD, HCLTech’s AI Factory, and both Accenture and TCS. GPU supply remains tight, so steering customers toward hardware they already run is an explicit part of the sales case. The database front showed up as well: NDB is certified with MongoDB Ops Manager, and NDB Time Machine delivers point-in-time recovery down to seconds.
What This Means for Enterprise AI Deployments
Three readings. Sovereignty and air-gapped operation have climbed from the footnotes to the headline: finance, healthcare, and public-sector buyers whose data may not leave the jurisdiction finally hold a complete infrastructure option pitched squarely at “run agents in place”, with Data Lens ransomware analytics reaching offline deployments too. Neocloud multitenancy shows the supply side of regional AI clouds assembling, handing enterprises a route between full self-build and the hyperscalers. And token usage monitoring now sits inside the platform layer itself — in step with the industry-wide turn from buying seats to managing traffic, since cost governance is exactly the hard problem of the next phase of enterprise AI.