Cornelis, a company building networking technology to improve communication between AI accelerators, announced on Monday that it has raised $205 million in a funding round led by IAG Capital Partners.
New product: Active Compute Fabric
The company unveiled a product called Active Compute Fabric, designed to address the considerable GPU time lost while waiting for data to arrive. Cornelis’s approach is a networking fabric that allows chips to process and send information concurrently, aiming to reduce idle periods and boost computational efficiency.
Open architecture and competition with Nvidia
Cornelis spun off from Intel in 2020. It positions itself as part of a new wave of AI infrastructure companies seeking to erode Nvidia’s market dominance by offering an open architecture: customers can use a variety of GPUs and accelerator hardware with Cornelis’s networking fabric. While Nvidia chips can technically run on other networking fabrics, they are typically optimized for Nvidia’s own software, which makes using Nvidia’s full GPU stack easier and more attractive for many customers.
Current status and plans
Cornelis has already begun shipping its product and is developing a next-generation version that it expects to release later this year. The newly raised $205 million is intended to support expansion of the networking technology and further product development.
Why this matters
As demand for AI workloads and high-performance computing grows, network efficiency becomes a key factor in both costs and performance. Cornelis’s solution targets GPUs’ idle waiting time directly, and its open approach may give customers more flexibility in choosing hardware and software components.
