Cornelis, a company specializing in networking solutions for artificial intelligence hardware, announced on Monday that it has secured $205 million in a funding round led by IAG Capital Partners. The investment supports the development of its Active Compute Fabric, a technology designed to address one of the most significant bottlenecks in AI processing: the time GPUs spend waiting for data.
Spinoff from Intel in 2020, Cornelis is part of a growing sector of infrastructure firms seeking to dismantle Nvidia’s monopoly on the AI chip market. While Nvidia’s GPUs are technically compatible with third-party networking systems, they are deeply optimized for Nvidia’s own software ecosystem, making its integrated stack the preferred choice for most customers.
Cornelis differentiates itself through an open architecture, allowing users to pair its networking fabric with a variety of GPU and accelerator hardware from different manufacturers. This approach aims to provide a more flexible and potentially cost-effective alternative to Nvidia’s closed system.
The Active Compute Fabric is engineered to enable chips to process and transmit information simultaneously, thereby minimizing wasted computational cycles. The company stated that it has already begun shipping its current product generation and is working on a newer version expected to launch later this year.
Finally, someone talking about reducing GPU idle time instead of just selling faster chips. Smart approach.
Does this actually work with AMD chips too? I am tired of vendor lock-in for my data center builds.
Twenty million dollars seems like a drop in the ocean compared to Nvidia’s R&D budget. They need more than just marketing.
Wait, a spinoff from Intel challenging Nvidia? That is quite the storyline. I want to see how this plays out.
Open architecture is exactly what the industry needs. Nvidia’s walled garden is getting expensive.