Nutanix and Netris Team Up on Multitenant AI Infrastructure for Neoclouds

The Neocloud Playbook Gets a Prefab Stack

Nutanix and Netris announced a strategic collaboration on October 21 at NVIDIA GTC Berlin to deliver a jointly engineered infrastructure stack built specifically for neoclouds and enterprises selling GPU cloud services. The pitch is simple: instead of assembling your own AI cloud from parts, buy the whole operating stack pre-integrated.

The combination pairs the Nutanix Cloud Platform, with its expertise in cloud infrastructure and compute orchestration, with the Netris NAAM platform, which handles network automation, abstraction, and multitenancy. Together they promise a scalable, secure, multitenant solution spanning the full infrastructure, from compute layers and data processing units down to the physical switches.

Why Multitenant AI Infrastructure Is Hard

Running AI workloads for multiple customers on shared GPU infrastructure is one of the nastier problems in cloud engineering. Tenants need strong isolation, because nobody wants their training run snooping on a neighbor's, but GPUs are expensive and idle GPUs are money on fire, so operators need to pack tenants densely. The industry has mostly solved the compute side with virtualization. The network side, where traffic between tenants must be separated at the hardware level, has lagged.

That is the gap the Nutanix-Netris stack targets. The joint solution promises hard isolation enforced on networking hardware between tenants, alongside the soft isolation of virtualization. Enterprises get to run business-critical AI workloads on multitenant GPU infrastructure with the kind of separation that used to require giving each customer their own physical gear. The combined offering also delivers automation and abstraction across compute and networking, so infrastructure can be provisioned and consumed through a cloud-like experience rather than by filing tickets with the network team.

The Neocloud Gold Rush

The timing is no accident. Demand for AI infrastructure has created a whole category of neoclouds, smaller GPU cloud providers racing to sell capacity that the hyperscalers cannot spare. Building a credible AI cloud used to mean hiring a team of infrastructure engineers and spending a year integrating compute, storage, and networking. A pre-validated stack compresses that to a purchasing decision, which is exactly what a neocloud racing to market wants to hear.

The economics favor it too. By simplifying how AI infrastructure is deployed and consumed across bare-metal and virtualized environments, neoclouds can scale more efficiently and improve margins as they grow. Faster time to revenue is the whole game when GPU capacity is the scarcest resource in tech and every quarter of delay is a quarter of someone else's bookings.

This is part of a broader pattern of AI infrastructure going prefab. NVIDIA now ships containerized data centers that arrive as shipping containers, and the industry is pouring billions into new capacity from Mexico to Johor. The Nutanix-Netris announcement is the software counterpart: the operating stack as a product, not a project.

What to Watch

Announcements at GTC are cheap; deployments are the proof. The questions that matter: which neoclouds actually standardize on the stack, how the hard-isolation claims hold up under independent security review, and whether the "cloud-like experience" survives contact with real enterprise networking teams and their very specific requirements.

But the strategic logic is sound. As AI's appetite for cloud capacity keeps outrunning supply, the winners will not just be whoever has the most GPUs. They will be whoever can turn GPUs into sellable, isolated, automated cloud services the fastest. Nutanix and Netris are betting that the fastest way to do that is to stop building it yourself.