In the race to build the next great AI cloud, the biggest enemy isn't a lack of capital or a shortage of chips. It’s the time it takes to turn a warehouse full of hardware into a functional, multi-tenant network. For a new "neocloud" provider, getting a GPU cluster online can take months of manual configuration.

That is a lifetime in the AI market, where every day a server sits idle is a day of wasted revenue. Netris, a network automation startup, is betting it can shrink that timeline from months to days. The company announced today it has raised $15 million in a Series A round led by Andreessen Horowitz to scale its platform, which automates the complex, hardware-level configuration required to run high-performance AI clusters.

The Hardware-Software Gap

For years, massive infrastructure players like AWS, Google, and Oracle built their own proprietary automation to manage the massive traffic loads required for cloud computing. They had the luxury of hiring armies of network engineers to bridge the gap between software-defined networking (SDN) and physical hardware.

Newer, smaller neocloud operators don't have that luxury. They are often forced to rely on traditional SDN tools that struggle to keep up with the sheer volume of data moving between GPUs.

"For AI, software is not okay, because the amount of traffic is so high, everything must be hardware accelerated," Netris CEO Alex Saroyan told TechCrunch. "You need something like SDN, but completely hardware accelerated. This is what we do."

Netris provides an abstraction layer that sits on top of network switches, allowing operators to manage their data center topology as if it were a single, programmable entity. By automating the configuration of every link, the platform allows neoclouds to support multi-tenancy—the ability to safely slice up hardware resources for different customers—without the manual overhead that usually plagues smaller operators.

Why a16z is Betting on Determinism

In an era where every startup is rushing to integrate generative AI into their product stack, Netris is taking a refreshingly old-school approach. The company explicitly avoids using AI to manage its network automation.

"AI is not deterministic, right?" Saroyan said. "Sometimes it likes to do things on its own. It’s good for creative work, but for changing many thousands of switch configurations, you don’t need to be creative. You need to be very persistent and repeatable."

This focus on deterministic, algorithmic automation has already won over some of the biggest names in the space. The company is currently live at more than 35 GPU clusters globally, managing a total of roughly one million GPUs. Its client list includes major players like Hewlett Packard Enterprise, Foxconn, and Lightning AI.

What This Means for Neoclouds

For the operators of these clusters, the value proposition is simple: speed to market. By removing the manual labor from network setup, Netris allows providers to start billing customers faster.

As part of the funding round, a16z partner Guido Appenzeller is joining the Netris board. The company plans to use the new capital to expand its engineering and sales teams, while adding support for a wider array of hardware vendors.

Key Takeaways

  • Solving the Bottleneck: Netris automates the complex, hardware-level network configuration that currently keeps new AI GPU clusters offline for months.
  • Hardware-Accelerated SDN: Unlike traditional software-defined networking, Netris focuses on hardware-accelerated automation to handle the massive traffic demands of modern AI training and inference.
  • Proven Scale: The platform is already managing approximately one million GPUs across 35 global clusters, with major enterprise clients including HPE and Foxconn.

The Road Ahead

With $15 million in the bank, the pressure is now on Netris to prove it can maintain its lead as the AI infrastructure market matures. The company’s next major hurdle will be hardware compatibility; as neoclouds experiment with a wider variety of chips and switch architectures, Netris must ensure its platform remains truly vendor-agnostic. If they succeed, they won't just be a tool for neoclouds—they will be the invisible plumbing that keeps the AI industry running.