AI infrastructure company Cornelis has raised $205 million to develop networking technology that helps AI chips communicate more efficiently, targeting a key challenge in large-scale AI computing.
IAG Capital Partners led the funding round. Cornelis also introduced Active Compute Fabric, a networking technology designed to reduce wasted GPU time caused by delays in moving data between chips.
The company says its approach allows AI processors to handle computation and data transfer at the same time, improving efficiency in systems that rely on large numbers of connected chips.
Cornelis targets Nvidia’s AI ecosystem
Cornelis is competing with Nvidia by offering an open networking architecture that can work with different GPUs and AI accelerators. Nvidia remains the leading supplier of AI chips, but its hardware is closely integrated with its own software and infrastructure tools.
The company, which spun out of Intel in 2020, is part of a growing group of AI infrastructure startups focused on improving individual parts of the computing stack rather than replacing Nvidia’s chips directly.
Cornelis has already started shipping its technology and is developing a new generation expected later this year, as demand grows for faster, more efficient AI data centres.
