Cornelis CN5000 deployed at Lawrence Livermore to support national security AI workloads

What happened at Lawrence Livermore National Laboratory?

The U.S. Department of Energy’s National Nuclear Security Administration (NNSA) has deployed Cornelis Networks’ CN5000 high-performance networking platform at Lawrence Livermore National Laboratory (LLNL) to power the 952-node “Lynx” supercomputing cluster. The announcement was made jointly by Cornelis Networks and the Department of Energy (DOE) on July 23, marking a significant step forward in modernizing the NNSA’s high-performance computing infrastructure under the Commodity Technology Systems (CTS-2) initiative.

The CN5000 platform will serve as the foundational network for the Lynx system, enabling faster and more reliable data throughput for mission-critical workloads under the Advanced Simulation and Computing (ASC) program. According to a statement by LLNL Senior Principal HPC Strategist Matt Leininger, the partnership aims to accelerate artificial intelligence (AI) and high-performance computing (HPC) simulations essential to national security operations.

The deployment was undertaken in partnership with Dell Technologies and follows a multi-year collaboration between Cornelis and the NNSA to develop a next-generation networking solution that meets the demands of large-scale simulation and AI environments.

Why is the CN5000 important for the NNSA’s simulation efforts?

The CN5000 platform represents a generational leap in network performance and scalability for supercomputing environments. Specifically designed to address challenges associated with compute underutilization, latency, and congestion, CN5000 enables high-throughput, lossless data transfer with sub-microsecond latency, critical for running demanding simulation and AI training models. It is based on Cornelis Networks’ Omni-Path architecture and integrates patented congestion management technology, advanced link-level control, and optimized scalability to ensure consistent performance across thousands of nodes.

In the context of the NNSA’s Advanced Simulation and Computing program, the CN5000 plays a critical role. The ASC program is tasked with developing simulation tools that can support the maintenance and modernization of the U.S. nuclear weapons stockpile without the need for explosive testing. These simulations require massive amounts of compute power and low-latency interconnects between thousands of nodes—requirements that CN5000 is engineered to meet.

Stephen Rinehart, Assistant Deputy Administrator for NNSA’s Advanced Simulation and Computing and Institutional R&D Programs, welcomed the launch of CN5000, indicating that the technology will have a continued impact across both the Commodity Technology and Advanced Technology computing systems managed by the DOE.

How does the CN5000 differ from previous systems?

Unlike conventional networking approaches that rely on TCP/IP or legacy interconnects, the CN5000 delivers an end-to-end scale-out fabric purpose-built for AI and HPC environments. It integrates seamlessly with existing compute infrastructures through support for industry-standard software, while also offering hardware innovations such as advanced SuperNICs (network interface cards), 48-port leaf switches, and 576-port modular Director Class switches.

The CN5000 product line also includes both air- and liquid-cooled components to address power and thermal constraints in modern data centers. In addition to hardware, Cornelis has released open-source host and network management software, which enables administrators to monitor and optimize network performance in real time.

This high level of integration and performance tuning was made possible through a three-year co-design effort between Cornelis and the DOE as part of the Next-Generation High Performance Computing Network (NG-HPCN) project. By embedding feedback and performance data from LLNL and other DOE sites into the design process, Cornelis ensured the CN5000 would meet the rigorous operational standards required for national security missions.

What is the significance of the Lynx cluster at LLNL?

The Lynx cluster is part of the Department of Energy’s broader mission to advance scientific computing for national security. It is one of the first systems to be deployed under the CTS-2 initiative, which seeks to refresh the DOE’s Commodity Technology Systems with modern, modular, and scalable compute clusters.

At 952 nodes, Lynx is a mid-scale cluster compared to larger exascale initiatives such as Frontier or Aurora. However, its focus is on enabling rapid deployment and testing of simulation codes, AI workflows, and pre-production research applications. These simulations are used by researchers at LLNL and other NNSA laboratories to ensure the safety, security, and effectiveness of the U.S. nuclear deterrent, and to support broader nonproliferation and defense programs.

LLNL’s involvement in the CTS-2 program highlights the lab’s leadership in simulation science, particularly in integrating AI methods with physics-based models. By deploying CN5000 as the backbone interconnect, the lab expects to reduce simulation runtimes, improve model fidelity, and support the development of next-generation simulation codes with AI-enhanced features.

What have officials said about the deployment?

In statements released as part of the announcement, both Cornelis Networks and DOE officials emphasized the strategic nature of the CN5000 deployment. Rob Hays, Vice President of Public Sector at Cornelis Networks, described the partnership as a major step in accelerating scientific discovery while reinforcing national security infrastructure.

Hays underscored Cornelis’ commitment to supporting the DOE in meeting mission objectives through advanced simulation tools. He stated that the company views CN5000 not merely as a commercial product, but as a co-engineered solution aligned with the computational needs of federal agencies working in sensitive domains like nuclear security, material science, and climate resilience.

Similarly, Matt Leininger of LLNL described the CN5000 as a critical component of the lab’s mission to advance AI and HPC. He noted that the platform’s congestion-avoidance and ultra-low latency capabilities would directly contribute to faster insights and higher productivity in simulation workflows.

What are the broader implications for U.S. computing infrastructure?

The deployment of CN5000 at LLNL comes amid a broader effort by the Department of Energy to modernize the nation’s scientific computing resources. Alongside exascale systems like Oak Ridge’s Frontier and Argonne’s Aurora, the DOE is investing in modular cluster systems under the CTS-2 program to support flexible, scalable compute environments.

These investments reflect a growing need to align national security, energy research, and climate science efforts with cutting-edge computational tools. CN5000’s role within this ecosystem is to enable efficient interconnect performance that supports both AI training and traditional physics-based simulations—fields that increasingly overlap as hybrid AI-physics models become more common.

Cornelis Networks has indicated that the CN5000 will be deployed in additional DOE labs and government agencies as CTS-2 and NG-HPCN rollouts continue. The company also plans to extend CN5000 capabilities through software updates and roadmap enhancements to support future innovations such as quantum simulation, neural architecture search, and energy-optimized computing.

How does this deployment fit within national security priorities?

At its core, the deployment of CN5000 reinforces the DOE’s commitment to maintaining technological superiority in computational science, which is vital to the NNSA’s core mission of nuclear deterrence and threat reduction. High-performance computing enables the simulation of nuclear physics scenarios that cannot be tested experimentally, ensuring both the safety and credibility of the U.S. stockpile.

As global security challenges become more complex, and as adversaries adopt advanced technologies, the role of AI-driven modeling and simulation will grow. CN5000 gives the DOE a foundational capability to meet this future—enabling more accurate forecasting, faster research cycles, and deeper understanding of critical defense systems.

From a public sector technology standpoint, the CN5000 deployment illustrates how strategic partnerships between government agencies and specialized hardware providers like Cornelis Networks can yield tailored solutions for domain-specific challenges. The success of this rollout is likely to influence future procurement strategies not only within the DOE but also in agencies involved in space, energy, and climate security.