Cornelis Networks Raises $205 Million to Challenge Nvidia in AI Networking

Cornelis Networks, a Pennsylvania-based AI infrastructure company spun off from Intel in 2020, has raised $205 million in new funding led by IAG Capital Partners, with participation from existing backers including Foresight Group, Intel Capital, KittyHawk Ventures, Ridgeline Partners and SQN Venture Partners. The round adds to the roughly $131 million the company had raised across its prior Series A, B and C rounds and multiple convertible notes since its founding.

Cornelis builds networking fabric that lets AI chips communicate with each other during large training and inference runs, a problem that becomes more acute as AI clusters scale into the tens of thousands of GPUs. The company's pitch to customers is an open architecture: rather than locking buyers into a single vendor's combination of accelerators and networking hardware, Cornelis' fabric is designed to work across a range of GPU and accelerator choices, built on open standards including UALink and ESUN. That stands in direct contrast to Nvidia, whose InfiniBand and NVLink networking technology is tightly integrated with its own GPU lineup and has become one of the company's most durable competitive moats.

The new capital will fund production of Cornelis' CN5000 and CN6000 switches, the hardware underpinning its newly announced Active Compute Fabric product, which targets a specific inefficiency in large AI clusters: GPUs sitting idle while they wait for data to arrive from other chips in the cluster. Cornelis says its fabric lets chips process and transmit information simultaneously rather than sequentially, aiming to close that idle-time gap as AI infrastructure operators push toward larger and more tightly coupled clusters.

The round also comes with a leadership signal. Lisa Spelman, who joined Cornelis as CEO in 2024 after a long run at Intel most recently leading the Data Center & AI Group's Xeon business, is steering the company through this scale-up phase, while co-founder Philip Murphy, the company's original CEO, has moved into the chief operating officer role.

For founders building anywhere in the AI infrastructure stack, Cornelis' raise is a reminder that the networking layer underneath AI compute, not just the chips themselves, remains an active and well-funded area of competition against Nvidia's dominant position. Open-standard alternatives at the networking layer give infrastructure buyers a genuine second option, a dynamic worth watching for any founder building on top of, or competing with, the major cloud AI providers.

The timing also lines up with a broader wave of AI data center buildout announcements this month, several of which appeared in AngelLinx Intelligence's own coverage over the past week, including large infrastructure raises from other AI compute providers. As hyperscalers and AI-native cloud companies push cluster sizes higher, the bottleneck increasingly shifts from raw chip supply toward how efficiently those chips can be wired together, which is precisely the market Cornelis is positioned to serve. Investors betting on Cornelis are effectively betting that no single company, including Nvidia, will be able to fully vertically integrate the entire AI compute stack, leaving room for independent networking specialists.

Founders working in AI infrastructure or adjacent hardware categories can find active investors through the match tool, browse AI/ML-focused investors in the directory, or check the live listing for current opportunities. Catch up on more coverage in the newsroom, or register on AngelLinx to start a raise.

AngelLinx Intelligence | angellinx.ai