Cornelis Networks

Overview

Cornelis Networks is the 2020 spinout of Intel’s Omni-Path business, building high-performance interconnect fabric for AI clusters. Its “Active Compute Fabric” architecture is pitched as a non-NVLink / non-InfiniBand path for large-scale training and inference deployments where independent studies (Meta + Harvard, Alibaba) report up to 32% of GPU-hours consumed by inter-node communication. As of September 2026, Cornelis is one of a small number of credibly-funded non-Nvidia interconnect vendors that Blackwell / Rubin cluster designers can benchmark against.

Timeline

• 2026-09-15-AI-Digest — Cornelis Networks closes a $205M round led by IAG Capital Partners, announced alongside a new “Active Compute Fabric” architecture and a strategic collaboration with Qualcomm. Series letter is not disclosed publicly; use of proceeds is CN5000 / CN6000 switch production. Load-bearing framing worth carrying: the bottleneck claim is not vendor spin — independent Meta + Harvard and Alibaba studies have reported up to 32% of GPU-hours consumed by inter-node communication on large training runs, so fabric-not-GPUs is increasingly the accepted bottleneck at cluster scale; whether Active Compute Fabric specifically is the answer is a Cornelis positioning claim, not a settled outcome. The Qualcomm wrinkle is worth logging: this is the second Qualcomm datacenter-AI angle in a week (see the Sep 8 AWS warrant deal on 2026-09-09-AI-Digest), signalling Qualcomm is broadening beyond edge silicon into the interconnect layer. Practitioner takeaway: ML infra teams evaluating Blackwell / Rubin cluster designs now have a credibly funded non-Nvidia interconnect stack to benchmark NVLink and InfiniBand against — three data points shy of a market, but the roadmap is legible enough to plan against. Reframe worth carrying: fabric bottleneck is industry-accepted, Cornelis's Active Compute Fabric is a proposed remedy — not the fix, not Cornelis solves the interconnect wall.

Key Developments

1. $205M Round + Qualcomm Strategic Collab + Active Compute Fabric Architecture (September 14, 2026): Cornelis Networks — the 2020 spinout of Intel’s Omni-Path business — closed a $205M round led by IAG Capital Partners, announced alongside a new Active Compute Fabric architecture and a strategic collaboration with Qualcomm. Series letter undisclosed; use of proceeds is CN5000 / CN6000 switch production. Load-bearing framing to carry: the bottleneck framing is not vendor spin — Meta + Harvard and Alibaba independently report up to 32% of GPU-hours consumed by inter-node communication on large training runs; whether Active Compute Fabric specifically is the answer is a Cornelis positioning claim, not a settled outcome. The Qualcomm wrinkle: second Qualcomm datacenter-AI angle in a week (see the Sep 8 AWS warrant deal), signalling Qualcomm is broadening beyond edge silicon into the interconnect layer. Practitioner takeaway: ML infra teams evaluating Blackwell / Rubin cluster designs now have a credibly funded non-Nvidia interconnect stack to benchmark NVLink and InfiniBand against — three data points shy of a market, but a legible roadmap. Reframe: fabric bottleneck is industry-accepted, Cornelis's Active Compute Fabric is a proposed remedy — not the fix, not Cornelis solves the interconnect wall. Covered in 2026-09-15-AI-Digest.

Related

See also: Qualcomm, NVIDIA, Amazon, MOC - AI Infrastructure, MOC - Major Companies.