Nutanix pushes bare‑metal AI play
Nutanix is sharpening its pitch as an 'enterprise AI control plane' and has launched NKP Metal to run Kubernetes directly on bare metal, signaling a push to attract customers who want high‑performance AI without a hypervisor layer. The vendor is pairing that message with storage and partner deals meant to lower migration friction for customers unsettled by VMware’s post‑acquisition changes. That combination positions Nutanix as a direct alternative for enterprises rethinking where Kubernetes and bare‑metal AI should run. (siliconangle.com (cloudnativenow.com)
Nutanix spent this week arguing that some artificial intelligence jobs should skip the virtual-machine layer entirely and run straight on the server, and it launched a product called NKP Metal on April 7 in Chicago to do exactly that. Nutanix announced it at its.NEXT 2026 conference, which ran April 7-9. (nutanix.com) (markets.businessinsider.com) Most corporate software still runs inside virtual machines, which are like separate apartments carved out inside one building. Nutanix is now saying some Kubernetes clusters for artificial intelligence training and edge systems work better when they get the whole building to themselves. (markets.businessinsider.com) (cloudnativenow.com) Kubernetes is the software that places and restarts containers, which are small packaged applications, across many machines. Bare metal means Kubernetes runs directly on physical servers instead of on top of a hypervisor, which is the software layer that slices one server into many virtual machines. (cloudnativenow.com) (markets.businessinsider.com) That missing layer is the whole sales pitch. Nutanix says running Kubernetes on bare metal can improve performance and flexibility for dense graphics processing unit clusters, edge deployments, and other jobs where customers do not want a hypervisor sitting between the application and the hardware. (markets.businessinsider.com) (sdxcentral.com) The company is not pitching this as a side tool. Nutanix says NKP Metal keeps the same operating model as its Nutanix Kubernetes Platform, so administrators can manage virtual machines and containers under one control system even when some workloads stay virtualized and others move onto raw hardware. (markets.businessinsider.com) (siliconangle.com) That is why Nutanix keeps using the phrase “enterprise artificial intelligence control plane.” At.NEXT, Chief Executive Rajiv Ramaswami framed Nutanix as the “operating model for AI factories,” while analysts covering the event said the company was trying to move from a cloud-platform vendor into the layer that governs cost, security, and deployment across enterprise AI systems. (siliconangle.com) (diginomica.com) The timing is not random. Broadcom closed its VMware acquisition on November 22, 2023, and Nutanix now has a public “Broadcom to Nutanix Migration Promotion” that offers some new customers up to one year of Nutanix licensing free, capped at 1,000 cores, if they commit to a minimum three-year term. (nutanix.com) Nutanix is also trying to remove one of the biggest reasons companies hesitate to move: storage. On April 7, Nutanix and NetApp said they would integrate NetApp storage with the Nutanix Cloud Platform and the Nutanix AHV hypervisor later in 2026, giving customers another way to keep existing storage investments while changing the compute stack. (netapp.com) The company is making the same argument with partners around networking and packaged infrastructure. Cisco said on April 7 that it is integrating Nutanix Enterprise AI, Nutanix Kubernetes Platform, and Nutanix Unified Storage into Cisco AI PODs and Cisco Unified Edge, with early access starting in April and general availability planned for June. (blogs.cisco.com) Put together, the message is simple: keep one management layer, move some workloads off the hypervisor, and make migrations less painful for VMware customers who do not want to rebuild everything at once. NKP Metal is the piece that lets Nutanix say it can handle both the old world of virtual machines and the newer world of graphics-processing-unit-heavy Kubernetes clusters on bare hardware. (markets.businessinsider.com) (nutanix.com)
