GITEX 2026: Netris Wants to Turn Massive GPU Clusters Into AI Clouds

Friday 23 October 2026 09:29

The artificial intelligence infrastructure race is often measured by the number of GPUs companies can deploy. Cloud providers are competing for the latest NVIDIA accelerators, governments are building sovereign computing capacity, and new AI factories are being planned at unprecedented scales.

But GPUs are only the beginning.

Once thousands of accelerators are installed inside an AI data center, operators still need to connect them, divide them between customers, isolate workloads, provision networks and continuously adjust infrastructure as demand changes.

That is the less visible challenge Netris is targeting at GITEX GLOBAL 2026.

As a Conference Headline Sponsor, the network automation company arrives in Dubai following rapid expansion across neoclouds, sovereign AI platforms and enterprise AI factories.

Its central proposition is simple: AI clouds require networking designed specifically around GPU infrastructure.

The Expensive GPU Utilization Problem

An AI accelerator generates value when it is actively performing useful work, making utilization one of the most important economic considerations in AI infrastructure.

A cloud operator may own thousands of GPUs worth hundreds of millions of dollars, but customers rarely need all of that capacity simultaneously.

One customer could require 256 GPUs, another 1,000, while a third might only need a small inference cluster. Their requirements can also change quickly.

The challenge is therefore not simply connecting machines. Operators need to divide shared physical infrastructure into isolated environments and reallocate computing resources without rebuilding the network each time.

Traditional networking processes can make that difficult.

Netris has built its platform around automating this layer.

NAAM Targets the Networking Challenges of AI Clouds

At the center of Netris’ strategy is NAAM, short for Network Automation, Abstraction and Multi-Tenancy.

The platform addresses three challenges that become increasingly important as GPU clusters expand.

Automation reduces the manual configuration required to provision switches, routing, load balancing, access controls and other network functions.

Abstraction provides cloud-like networking capabilities without requiring users to understand the physical hardware underneath.

Multi-tenancy enables different customers and workloads to share infrastructure while remaining isolated.

Together, these capabilities allow operators to treat physical GPU infrastructure more like a programmable cloud.

Instead of manually configuring individual devices for every new customer, operators can define the required environment and allow the platform to translate those requirements into the underlying network configuration.

From Physical Switches to a Virtual AI Cloud

Consider an AI provider operating 10,000 GPUs.

One customer may want a private environment containing 500 accelerators, another may require 2,000 and a third could need 200.

Each environment may require different networking, security policies, IP addresses and connectivity.

More importantly, customer traffic must remain isolated.

Netris creates virtual network environments on top of shared physical infrastructure while leaving the underlying switches and links in place.

The approach resembles the way public cloud platforms have abstracted conventional computing infrastructure.

Developers using virtual private clouds generally do not configure physical switches inside hyperscale data centers. Instead, they request resources through software.

Netris is applying a similar concept to GPU clouds.

Multi-Tenancy Is Becoming Critical to AI Economics

This is where Netris’ technology becomes particularly relevant from a commercial perspective.

Dedicated GPU clusters can become inefficient when customers do not continuously use all the capacity allocated to them.

Multi-tenancy allows operators to serve multiple customers from a common hardware pool.

However, sharing infrastructure also introduces security and governance requirements.

A bank running an AI workload may require strict network isolation. A government operating sovereign AI infrastructure may need clear separation between different workloads, while an AI developer training a proprietary model may require similarly strong controls.

Netris implements isolation at the networking hardware level across bare-metal, virtual-machine and container environments.

The objective is to make shared AI infrastructure operate as though each tenant has its own network.

If that approach works at scale, operators could potentially improve GPU utilization without compromising workload isolation.

AI Networks Are Becoming More Complex Than Ethernet

Modern AI clusters rarely rely on a single networking technology.

Infrastructure can include conventional Ethernet, NVIDIA Spectrum-X, InfiniBand, storage networks and management networks. New architectures also introduce technologies such as NVIDIA BlueField DPUs and NVLink.

Each layer can have different configuration and operational requirements.

Netris is attempting to provide a unified automation and control layer across these different network fabrics instead of forcing operators to manage each infrastructure component separately.

Its platform supports Ethernet, InfiniBand, NVIDIA Spectrum-X, BlueField DPUs and NVLink environments.

That breadth becomes increasingly important as AI factories move toward rack-scale and eventually data-center-scale architectures.

BlueField DPUs Push Tenant Isolation Deeper Into AI Infrastructure

During 2026, Netris expanded its network automation and multi-tenancy capabilities to NVIDIA BlueField DPUs.

A DPU, or data processing unit, handles infrastructure functions that would otherwise consume CPU resources.

In AI clouds, BlueField can play an important role in networking, security and workload isolation.

Netris extends its control plane across Ethernet switches and BlueField DPUs, allowing operators to manage tenant isolation across multiple layers of the infrastructure.

This matters because customer boundaries no longer exist only at the top-of-rack switch.

As AI systems become increasingly integrated, workload isolation must follow the workload deeper into the computing architecture.

Kubernetes Connects AI Workloads With the Physical Network

Containers introduce another layer of complexity.

Kubernetes can effectively orchestrate applications and computing resources, but the physical network underneath a large GPU cluster still needs to be configured correctly.

Netris has been working to bridge that gap.

Its integration with Mirantis connects Kubernetes orchestration with network automation, allowing operators to coordinate computing and networking instead of managing them as separate operational domains.

The goal is a more automated approach to building AI clouds.

When Kubernetes provisions a workload, the required network environment can also be created.

This reflects the broader movement toward software-defined operations across AI infrastructure, where GPUs, containers and networks need to work as a coordinated system.

NVIDIA DSX Air Brings Networking Into the Digital-Twin Process

Netris has also expanded its integration with NVIDIA DSX Air during 2026.

DSX Air provides a digital environment for designing and validating AI infrastructure before physical deployment.

Netris brings its network automation and multi-tenancy capabilities into that environment.

The approach can be particularly valuable for large AI clusters.

A configuration mistake discovered after thousands of GPUs and network components have already been installed can be extremely costly.

A digital twin allows infrastructure teams to model, test and validate designs before physical deployment.

For companies building sovereign AI platforms and neocloud infrastructure, that could help reduce deployment risks and shorten time to service.

A 35,000-GPU Cloud Shows the Challenge at Scale

Netris is already operating beyond small-scale demonstrations.

Lightning AI standardized on the platform for network automation and multi-tenancy across infrastructure containing more than 35,000 NVIDIA GPUs in six data centers.

The deployment illustrates why automation becomes increasingly important as AI infrastructure expands.

Managing networking across a handful of servers is possible manually.

Managing tens of thousands of accelerators distributed across multiple facilities and customers represents a fundamentally different operational challenge.

TELUS has also used Netris as part of the network automation and multi-tenancy foundation for its sovereign AI infrastructure in Canada.

In Taiwan, Foxconn-backed Visionbay.ai selected Netris for what it describes as the country’s largest GPU cluster and AI supercomputing center.

These deployments give Netris a significant message for GITEX: its technology is already being applied to production AI infrastructure.

Netris Is Growing Alongside the Neocloud Boom

Netris’ growth reflects the rapid development of the neocloud market.

The company raised $15 million in Series A funding in June in a round led by Andreessen Horowitz.

At the time, Netris reported 800% annual recurring revenue growth and more than 35 live AI cluster deployments during the previous 12 months.

By September, when the company announced its partnership with World Wide Technology, Netris said annual recurring revenue growth had reached 1,000%, with more than 45 live deployments across neoclouds, sovereign AI providers and enterprise AI factories.

That expansion is closely linked to the rise of neocloud providers.

These companies specialize in GPU computing for AI rather than offering the broad range of services associated with traditional hyperscale cloud providers.

Their commercial challenge is straightforward: acquire expensive accelerators, deploy them quickly and keep as much of that capacity productive as possible.

Networking is central to all three objectives.

WWT Takes Netris Deeper Into the AI Factory

Netris’ September partnership with World Wide Technology represents another important step in its expansion.

WWT is standardizing on Netris as the network automation, abstraction and multi-tenancy layer for its end-to-end AI factory offering.

That places Netris within a broader infrastructure proposition covering compute, storage, facilities and networking.

The significance extends beyond another technology partnership.

It moves Netris closer to the process of building complete AI factories.

Customers planning large GPU deployments can receive computing and physical infrastructure while Netris provides the programmable networking layer needed to turn those systems into usable cloud capacity.

That could be particularly relevant in the Gulf, where organizations are increasingly building new AI infrastructure from the ground up.

Sovereign AI Makes Networking More Important

Sovereign AI is expected to remain a major theme around GITEX GLOBAL 2026.

Governments are increasingly seeking domestic computing capacity capable of processing sensitive data and supporting local AI development.

Building that capacity requires more than purchasing GPUs.

A sovereign AI provider may need to serve government departments, universities, national companies, startups and strategic industries using shared infrastructure.

These organizations require isolation, governance and predictable access to computing resources.

Multi-tenancy therefore becomes part of the sovereignty equation.

The physical hardware may be located within national borders, but the infrastructure still needs software capable of determining exactly who can access which resources.

The Middle East Could Be a Major Market for Netris

The Gulf is investing heavily in the types of infrastructure Netris is designed to manage.

Large AI clusters, sovereign cloud platforms, data centers and GPU-as-a-service offerings are increasingly becoming part of national technology strategies.

At the same time, the economics of AI infrastructure are becoming more demanding.

GPUs are expensive, power availability is constrained and new data-center capacity takes time to build.

Leaving expensive accelerators underused because networking cannot be provisioned quickly is therefore increasingly difficult to justify.

Netris does not create additional GPUs.

Instead, its technology aims to make existing GPU capacity easier to divide, provision and keep productive.

What to Expect From Netris at GITEX GLOBAL 2026

Netris has confirmed its participation in GITEX GLOBAL 2026, although it has not disclosed a complete schedule of Dubai-specific product announcements.

Its recent developments provide a strong indication of the areas likely to attract attention.

NAAM should remain central to the company’s story as network automation and multi-tenancy become increasingly important for commercial GPU clouds.

The expansion into BlueField DPUs demonstrates how tenant isolation is moving deeper into AI infrastructure.

The NVIDIA DSX Air integration shows how network design can become part of the digital-twin process before an AI factory is physically built.

Meanwhile, the WWT partnership demonstrates Netris’ movement from standalone networking software toward complete AI factory deployments.