Cornelis, spun off from Intel in 2020 to build networking tech that helps AI chips communicate more effectively, raised $205M led by IAG Capital

You can now use networking hardware that allows your AI accelerators to process data more efficiently.

Cornelis, an AI infrastructure company that spun off from Intel in 2020, has secured $205 million in funding led by IAG Capital. The company has also launched its Active Compute Fabric, a networking technology designed to improve the communication efficiency between AI chips and reduce GPU idle time. This technology aims to address the bottleneck of data transfer, allowing chips to process and transmit information concurrently. Cornelis positions its offering as an open architecture solution, enabling customers to integrate various GPU and accelerator hardware, thereby challenging Nvidia's dominant position in the market. Nvidia's ecosystem is currently optimized for its proprietary software, making it more convenient for users to adopt its full hardware stack. Cornelis has already begun shipping its product and plans to release an updated version later this year.

Cornelis's $205 million funding round and the launch of its Active Compute Fabric signal a growing challenge to Nvidia's entrenched AI hardware dominance. By offering an open architecture that supports diverse accelerators, Cornelis is attempting to commoditize the AI networking layer, a critical component currently tightly integrated with Nvidia's proprietary stack. This move could empower AI infrastructure builders to avoid vendor lock-in and potentially lower costs by mixing and matching hardware from different suppliers.

The company's strategy focuses on optimizing data flow to eliminate GPU waiting times, a significant source of inefficiency in large-scale AI training and inference. As AI models become more complex and data-intensive, the performance of the underlying networking fabric will become increasingly crucial. Cornelis's success could encourage further innovation in open AI infrastructure, potentially leading to a more fragmented and competitive market for AI accelerators and their supporting components.