Netris raises $15 million to automate AI data center networking

Netris has raised $15 million in a Series A funding round from Andreessen Horowitz to expand its platform. The startup provides software that runs on network switches to help neocloud operators—which are small data center businesses providing AI inference and training services—reduce the time it takes to go live. The platform automates setup, configuration, and operations, addressing a critical bottleneck for these infrastructure providers. The software also helps these operators manage multi-tenancy, which is an architecture where a single instance of software serves multiple customers.

The company has spent eight years developing its hardware-accelerated approach. Alex Saroyan, Netris’ CEO, explained that traditional software-defined networking (SDN) is falling short for AI workloads because it is a software-based technology. “For AI, software is not okay, because the amount of traffic is so high, everything must be hardware accelerated. So you need something like SDN, but completely hardware accelerated. This is what we do, and this is what we’ve been doing for eight years,” Saroyan said. He noted that GPU cluster operators must make configuration changes to every link daily, which traditional software-defined networking cannot handle efficiently because it operates purely in software rather than being hardware accelerated.

Netris is currently live at more than 35 GPU clusters around the world, which are groups of interconnected GPUs used for computing. In total, the startup’s software manages about a million GPUs. Two years ago, chipmaking giant Nvidia recommended Netris to several customers after a technology demonstration. Today, the startup’s customer base includes:

• Lightning AI
• Foxconn
• Visionbay
• Hewlett Packard Enterprise
• TensorWave
• Telus

Saroyan emphasized that the platform relies on deterministic algorithms rather than generative AI to manage these configurations. He explained that Netris started developing its algorithm long before the AI boom because the team understood the challenge early on. Because AI is not deterministic, Saroyan noted that it is unsuitable for changing thousands of switch configurations. While AI is useful for creative work, managing complex network switches requires a highly persistent and repeatable process rather than creativity. To support the company’s future scaling, Andreessen Horowitz partner Guido Appenzeller is joining the Netris board of directors.

Why it matters

As the AI boom drives demand for new data centers, neocloud operators face a bottleneck in network configuration. Netris provides the automation layer to help these smaller players compete with hyperscalers.