Cornelis, an Intel spinout developing networking chips, has said a U.S. supercomputer is using its chips for nuclear work.
Lawrence Livermore National Laboratory said Tuesday that it has tapped Cornelis chips to connect 952 computers in its new “Lynx” system. The Lynx system is part of a $70 million program at the three labs working to develop and maintain the nation’s nuclear weapons with extremely accurate computer simulations of nuclear reactions, one of the most demanding tasks in the entire computing industry. This program is aimed at building workhorse supercomputers out of standard, off-the-shelf components from the computing industry.
READ: Intel to invest $15 million in chip startup SambaNova (April 2, 2026)
“We are excited to see the Cornelis CN5000 400G network come to life at Lawrence Livermore National Laboratory,” said Matt Leininger, Senior Principal HPC Strategist at LLNL. “The collaboration between NNSA’s ASC program and Cornelis has been rooted in a shared commitment to advancing high-performance computing. Lynx reflects the results of that public-private R&D investment and will support the modeling, simulation and analysis capabilities that underpin the modern NNSA complex.”
According to Businesswire, Lynx is a key computing infrastructure investment for the National Nuclear Security Administration (NNSA). “Lynx represents an important milestone in NNSA’s work to evaluate and deploy next-generation high-performance computing technologies for mission use,” said Stephen Rinehart, Assistant Deputy Administrator for the NNSA’s Office of Advanced Simulation and Computing. “The system builds on NNSA’s Next-Generation High Performance Computing Network effort and strengthens the computing ecosystem supporting future ASC workloads.”
READ: Intel to retain networking and communications unit in ongoing restructuring (December 4, 2025)
Cornelis had spun out of Intel in 2020. The company is currently developing a networking technology called Omni-Path that seeks to challenge rival networking chips from Nvidia
and Broadcom for applications where a computing problem is so large that it must be spread out over many different computers.
The Cornelis CN5000 fabric used for Lynx utilizes the Omni-Path architecture. This provides low-latency, lossless and congestion-free communication to maximize compute performance and efficiency for today’s HPC and AI workloads.
One of the features of Cornelis chips is traffic routing technology that, for example, recognizes when it may be faster to send data to a computer that’s further away if all nearby computers are clogged with network traffic. “You might drive a mile longer, but you get there 10 minutes faster because you avoided the stadium traffic from the FIFA World Cup,” Cornelis CEO Lisa Spelman said, explaining the concept.