Key facts
- AI infrastructure company Cornelis raised $205 million in a funding round led by IAG Capital Partners.
- Cornelis announced its Active Compute Fabric, a network technology designed to reduce GPU idle time by allowing chips to process and send information simultaneously.
- The company aims to challenge Nvidia's market dominance by offering an open architecture that supports various GPU and accelerator hardware.
- Cornelis's CN5000 networking product is already shipping, and the CN6000 is expected to be more broadly available in Q4 2026.
- The company spun off from Intel in 2020 and has a history in high-performance computing networking.
Cornelis, an AI infrastructure company specializing in networking technology, announced on Monday that it has secured $205 million in a funding round led by IAG Capital Partners. The company aims to challenge Nvidia's dominance in the AI chip market by providing an open architecture for networking fabrics.
Cornelis's newly launched Active Compute Fabric architecture is designed to improve the efficiency of AI accelerators by processing and transmitting data simultaneously, thereby reducing the time GPUs spend waiting for data. This technology aims to alleviate congestion in large AI clusters. The company's CN5000 switch, offering 400 Gbps throughput, is already shipping, while the CN6000, supporting 800 Gbps, is slated for broader availability in the fourth quarter of 2026.
Spun off from Intel in 2020, Cornelis leverages open standards such as Ultra Ethernet and UALink, positioning itself as an alternative to Nvidia's proprietary interconnect systems. The company has a history in high-performance computing, with its networking technology used in institutions like Lawrence Livermore National Laboratory's Lynx supercomputer. A strategic partnership with Qualcomm Technologies further reinforces its open-standards approach.