High-performance networking solutions provider Cornelis Networks has announced "Active Compute Fabric™," an open architecture that supports both scale-up and scale-out networking while integrating programmable computing capabilities directly into the fabric. With this new architecture, the company is making a full-scale entry into the scale-up networking market, expanding beyond its traditional domain of scale-out.
Cornelis Networks also announced that it has secured $205 million in funding to accelerate business expansion and bring its next-generation products to market. The raised capital will be allocated toward scaling production capacity, strengthening customer partnerships, and accelerating market deployment plans.
As AI models grow increasingly massive, communication overhead has emerged as a critical bottleneck that reduces GPU utilization rates. Active Compute Fabric integrates lossless transport, in-fabric acceleration, and programmable computing to offload data processing to the network side. By doing so, it aims to reduce the time accelerators like GPUs spend waiting for data and improve the overall economics of the system.
At the AI Infra Summit, Cornelis announced a collaboration with Qualcomm Technologies. Sharing a common vision regarding the role of networking in efficient AI infrastructure, the two companies are promoting an integrated approach to rack-scale AI systems. Cornelis is currently shipping its "CN5000" in volume production, has begun sampling its next-generation "CN6000," and aims to capture market share in a sector projected to reach $55 billion by 2030.