Cornelis Networks raised $205 million in a round led by IAG Capital Partners, the AI infrastructure company said, to scale production of its CN5000 and CN6000 switches and expand its new Active Compute Fabric.
Active Compute Fabric targets a specific inefficiency: GPUs sitting idle while they wait for data to arrive. The fabric lets chips process and send information at the same time instead of waiting on each transfer, according to TechCrunch and SiliconANGLE.
Cornelis, which spun off from Intel in 2020, sells an open architecture that works across different vendors' GPUs and accelerators, positioning it against Nvidia's networking stack, which is built to work best with Nvidia's own chips, TechCrunch reported. The new fabric was announced alongside a strategy collaboration with Qualcomm, SiliconANGLE reported.
For teams buying AI networking gear, Cornelis's pitch is portability: a fabric that doesn't lock a data center into one chip vendor. Whether that wins budget away from Nvidia's integrated stack will depend on whether the performance gap closes as fast as the funding does.