The rush to build AI infrastructure has created a practical problem: owning the right hardware is only the beginning. For data center operators serving AI inference and training, GPUs, network switches, and storage still have to be configured, operated, and adapted for customers before revenue can begin.
Netris is targeting that gap. The network automation startup says its software can reduce the time it takes for AI neoclouds to launch services by automating setup, configuration, and ongoing operations across GPU clusters.
Why AI Neoclouds Need Faster Network Setup
AI demand has pushed more companies toward the data center business, but bringing a facility online is not simple. Even after securing GPUs and other infrastructure, operators face months of work before they can provide cloud-computing services for AI training and inference.
That delay matters because idle GPUs are expensive assets. The longer a neocloud takes to reach customers, the longer its infrastructure sits unused while costs continue to build.
Large infrastructure companies have historically had an advantage here. Operators such as Equinix, NTT, Digital Realty, Oracle, Microsoft, AWS, and Google have been able to solve networking, configuration, and multi-tenancy challenges with large engineering teams or their own automation systems.
Smaller neocloud businesses usually do not have that same depth of internal engineering capacity. That is where Netris is positioning itself: as a ready-made automation layer for the networking work that has traditionally required deep infrastructure expertise.
What Netris Automates
Netris provides software that runs on network switches, along with a platform that connects to those switches. Its system is designed to automate data center network setup, configuration changes, and operations for GPU cluster operators.
The company also provides network abstraction. In plain terms, that means operators can change hardware configurations as needed without having every operational detail tied tightly to one fixed setup.
Another core part of the platform is hardware-layer isolation of servers and resources. That matters for multi-tenancy, because neoclouds need to serve multiple customers while keeping their workloads and resources separated.
For AI infrastructure, Netris CEO Alex Saroyan argues that traditional software-defined networking is not enough. He told TechCrunch that GPU cluster operators need to make configuration changes to every link, every day, and that AI traffic levels require hardware acceleration rather than a purely software approach.
Saroyan described Netris as something like software-defined networking, but fully hardware accelerated. He said the company has been working on that problem for eight years.
A Vendor-Agnostic Approach for GPU Clusters
According to Saroyan, the Netris platform is vendor-agnostic. It is compatible with networking equipment and standards used in data centers, including environments built around Nvidia and AMD servers.
That compatibility is important because neocloud operators may need flexibility across hardware choices. A platform that works across vendors can be more useful to operators that are scaling quickly, changing configurations, or supporting different customer requirements.
The company has already gained traction inside the AI infrastructure market. Two years ago, Nvidia saw a demo of Netris technology and recommended the company to several customers.
Netris is now live at more than 35 GPU clusters around the world, representing about a million GPUs total. Customers and operators named in the source include Lightning AI, Foxconn, Visionbay, Hewlett Packard Enterprise, TensorWave, Telus, and others.
The $15 Million Series A
Netris has raised $15 million in a Series A round from Andreessen Horowitz. TechCrunch reported the financing as an exclusive.
a16z partner Guido Appenzeller is joining the company’s board. With the new funding, Netris plans to hire more engineers and sales staff, add support for more hardware vendors, and build more functionality into its algorithm.
One notable detail is what Netris says it is not using: AI. Saroyan said the company relies on algorithms it had already developed for automation, configuration, and operations.
The reasoning is straightforward. For large-scale switch configuration, Saroyan said the goal is not creativity, but persistence and repeatability. In an environment where many thousands of switch configurations may need to change, deterministic behavior is central to the value proposition.
What This Signals for AI Infrastructure
The Netris funding round reflects a broader infrastructure reality behind the AI boom. The market does not only need more GPUs; it also needs the operational systems that make those GPUs usable for customers.
For neoclouds, the business challenge is speed. Hardware that is not yet configured for cloud-computing services cannot serve AI training or inference customers, which turns technical delay into commercial drag.
Netris is betting that network automation, hardware acceleration, abstraction, and multi-tenancy support can become a core layer for these operators. If its platform can shorten the path from GPU cluster installation to live customer service, it addresses one of the less visible bottlenecks in the AI infrastructure buildout.
The company’s next phase will center on execution: expanding the team, broadening hardware vendor support, and extending the algorithm that underpins its automation platform.