What happened
Cornelis, an Intel spinoff building AI networking fabric, raised $205 million led by IAG Capital Partners and unveiled its Active Compute Fabric platform.
AI infrastructure startup Cornelis has raised $205 million in a funding round led by IAG Capital Partners to scale its open networking technology for AI clusters. The company builds fabric that helps AI chips communicate more efficiently, letting them compute and transfer data at the same time instead of idling.
Spun off from Intel in 2020, Cornelis is positioning itself as an open alternative to Nvidia's tightly integrated stack. While Nvidia GPUs can run on third-party fabrics, they are optimized for Nvidia's own software, which keeps customers inside its ecosystem.
Its new platform, Active Compute Fabric, is central to that pitch. The CN5000 switch, already shipping, delivers 400 Gbps throughput, while the next-generation CN6000 will hit 800 Gbps with multi-protocol support and is expected broadly in Q4 2026.
The round is part of a larger wave of investment in Nvidia alternatives, alongside Nexthop AI's $500M and Eridu's $230M raises. Cornelis estimates that in a 100,000-GPU cluster, nearly half of GPU time is wasted waiting for data — about $1.68 billion and 500 GWh per year.
How Cornelis got here
From Intel origins to an open AI fabric challenger.
- 2020 — Cornelis spins off from Intel to focus on high-performance networking.
- 2026-09-14 — Announces $205M round led by IAG Capital and unveils Active Compute Fabric.
- Now — CN5000 shipping at 400 Gbps; CN6000 at 800 Gbps targeted for Q4 2026.
- Next — Qualcomm partnership showcase planned at AI Infra Summit.
What to watch next
CN6000 delivery in Q4 2026 and Qualcomm showcase will test whether open standards can convert Nvidia-locked customers. Field performance and pricing vs Nvidia interconnect will decide adoption.
Why it matters for me
AI data centers get a faster, open switch option (CN5000 shipping at 400 Gbps, CN6000 at 800 Gbps in Q4 2026) that works across vendors instead of locking into Nvidia.
Money
Business Executives · Developers
Data-center operators could reclaim billions in wasted GPU time — Cornelis models ~$1.68B/year lost per 100K-GPU cluster when chips idle waiting for data.
Work
Developers · Business Executives
Teams building AI clusters get a vendor-neutral networking option that works across GPUs and accelerators, reducing lock-in to Nvidia's stack.
Daily Life
General Public
More efficient fabric could cut massive power waste — ~500 GWh per 100K-GPU cluster — relevant as AI energy use faces public scrutiny.
What to remember
Open networking could save billions in wasted GPU time.
Intel spinoff Cornelis raises $205M to take on Nvidia with open 800 Gbps AI networking fabric.
Verified sources (4)
AI infrastructure company Cornelis raises $205M to chip away at Nvidia's dominance
↗Cornelis raises $205M to challenge Nvidia's dominance in AI networking
↗Cornelis Raises $205M for AI Networking, Unveils Active Compute Fabric with Qualcomm
↗Cornelis raises $205 million to challenge Nvidia with open AI networking
↗Claims and linked sources
7 claimsCornelis, an AI infrastructure company creating networking technology to help AI chips communicate more effectively, raised $205 million in a round led by IAG Capital Partners.
Cornelis launched Active Compute Fabric, a networking technology designed to reduce GPU time wasted waiting for data by letting chips process and send information at the same time.
Cornelis is an Intel spinoff from 2020 and competes with Nvidia by offering an open architecture that works with a variety of GPU and accelerator hardware, not tied to one vendor.
Nvidia GPUs can technically run on other networking fabrics but are optimized to run on Nvidia's own software, making it easier for customers to use Nvidia's full stack and reinforcing its dominant position.
Cornelis' CN5000 switch is already shipping at 400 Gbps, while the CN6000 supports multiple protocols at 800 Gbps and is expected to be broadly available in Q4 2026.
Cornelis supports open standards including Ultra Ethernet and UALink, both backed by industry consortiums seeking alternatives to Nvidia's proprietary interconnect stack.
Cornelis announced a strategic partnership with Qualcomm Technologies, with the two planning to showcase the collaboration at the AI Infra Summit.