Netris Raises $15M from a16z to Accelerate AI Neocloud Deployment

As the race for GPU supremacy intensifies, the bottleneck for new AI cloud providers has shifted from acquiring chips to configuring the complex networks required to run them. Netris, a network automation startup, has secured $15 million in Series A funding led by Andreessen Horowitz (a16z) to solve this critical infrastructure challenge.

Solving the Neocloud Deployment Bottleneck

The rise of "neoclouds"—specialized cloud providers focusing on AI inference and training—has created a massive demand for rapid data center deployment. Traditionally, setting up a data center to support high-performance computing can take months, leaving expensive GPUs sitting idle while engineers struggle with manual configurations.

Netris addresses this by providing a specialized software platform that runs directly on network switches. Their technology automates the setup, configuration, and day-to-day operations of data centers. Crucially, the platform offers network abstraction and hardware-layer resource isolation, enabling multi-tenancy. This allows neocloud operators to serve multiple customers securely on the same hardware, a feat that previously required massive engineering teams like those at AWS or Google.

Hardware Acceleration vs. Traditional SDN

A key technical differentiator for Netris is its approach to networking speed. CEO Alex Saroyan notes that traditional Software-Defined Networking (SDN) is insufficient for the massive traffic demands of AI workloads. Because SDN is primarily a software-based technology, it cannot match the required throughput.

Netris provides what Saroyan describes as "hardware-accelerated SDN." By moving the intelligence closer to the hardware, the platform can handle the extreme data volumes of AI clusters without the latency penalties of software-only solutions. This vendor-agnostic approach ensures compatibility with both Nvidia and AMD server ecosystems, making it a versatile tool for diverse hardware stacks.

Reliability Through Deterministic Algorithms

Interestingly, despite operating in the AI infrastructure space, Netris does not use AI to manage its networks. Saroyan explains that for managing thousands of switch configurations, "creativity" is a liability. AI is non-deterministic, meaning it can produce unpredictable results—a risk no data center operator can afford.

Instead, Netris relies on highly persistent and repeatable algorithms developed over the last eight years. This deterministic approach ensures that network changes are predictable and exact. The efficacy of this method is evidenced by its current scale: Netris is already live at more than 35 GPU clusters globally, managing approximately one million GPUs for major players including Lightning AI, Foxconn, HPE, Tensorwave, and Telus.

Scaling for the Future of AI Infrastructure

With the new $15 million infusion and the addition of a16z partner Guido Appenzeller to the board, Netris plans to aggressively expand. The company aims to increase its engineering and sales headcount, support a wider array of hardware vendors, and deepen the functionality of its core automation algorithms. As the AI industry moves from experimental clusters to massive-scale production environments, the ability to deploy "turnkey" networking will be a decisive factor for the next generation of cloud providers.

Key Takeaways

  • Rapid Deployment: Netris automates the complex configuration of GPU clusters, significantly reducing the time it takes for neoclouds to go live and start generating revenue.
  • Hardware-Accelerated Networking: Unlike traditional SDN, Netris provides hardware-level acceleration to handle the extreme traffic demands of AI training and inference.
  • Proven Scale: The platform is already operational across roughly one million GPUs, serving industry leaders like Foxconn, HPE, and Lightning AI.