AI infrastructure company Cornelis raises $205M to chip away at Nvidia’s dominance

Cornelis, a networking technology company spun out of Intel in 2020, said on Monday it raised $205 million in a funding round led by IAG Capital Partners and introduced a product called Active Compute Fabric. The fabric is designed to reduce GPU idle time by enabling chips to process and transmit data concurrently; Cornelis has begun shipping the product and plans a next-generation release later this year.

By AI NewsroomPublished about 1 hour agoUpdated about 1 hour ago0 views
AI infrastructure company Cornelis raises $205M to chip away at Nvidia’s dominance

Why It Matters

The technology targets a common inefficiency in AI deployments—GPUs waiting for data—which could improve hardware utilization. Coupled with an open architecture that supports multiple GPU and accelerator vendors, Cornelis’s approach aims to offer customers an alternative to the tightly integrated Nvidia software and hardware stack.

Key Facts

  • Funding raised: $205 million
  • Lead investor: IAG Capital Partners
  • Product announced: Active Compute Fabric
  • Product goal: Reduce GPU idle time by letting chips process and send data at the same time
  • Company origin: Spun off from Intel in 2020

Cornelis announced a $205 million funding round led by IAG Capital Partners and introduced a new networking product called Active Compute Fabric. The company said the fabric is intended to address a common bottleneck in large-scale AI systems where GPUs sit idle while waiting for data to arrive.

Active Compute Fabric is built to let chips perform computation and data transmission simultaneously, a design aimed at improving overall GPU utilization. Cornelis positions the technology as a way to reduce wasted GPU time rather than focusing solely on raw compute performance.

The company offers an open architecture that supports a variety of GPUs and accelerator hardware, differentiating itself from Nvidia’s more vertically integrated approach. While Nvidia GPUs can run on other networking fabrics, their software optimizations make Nvidia’s own stack easier for many customers to deploy, which has contributed to Nvidia’s strong market position. Cornelis is among a group of infrastructure startups attempting to offer alternatives to that dominant model.

Cornelis said it has begun shipping its current product and is developing a new generation expected later this year. The company, which spun out of Intel in 2020, is using its networking fabric and open approach to give customers more choice in how they assemble AI hardware and software stacks.

Keep Reading