Cornelis Networks has secured $205 million in a funding round led by IAG Capital Partners. The company plans to use these funds to increase production of its CN5000 and CN6000 network switches and to accelerate the deployment of its new Active Compute Fabric architecture.
The Active Compute Fabric is an open architecture designed to support scale-up and scale-out networks for AI workloads. It integrates programmable compute directly into the network, allowing it to perform tasks like assembling KV cache data and compressing gradients while data is in transit.
According to the company, this approach aims to reduce network traffic by up to 50% by minimizing the time GPUs spend waiting for data synchronization. The architecture utilizes open standards including Ethernet, UALink, and Ultra Ethernet to ensure compatibility with existing compute systems.
Alongside the funding announcement, Cornelis Networks revealed a strategic collaboration with Qualcomm Technologies. Both companies intend to focus on the development of networking solutions for rack-scale AI data centers, treating the network as a primary design component rather than a passive infrastructure element.
