Lenovo Adds Cornelis CN5000 to EveryScale Platforms to Eliminate AI/HPC Network Bottlenecks

-Cornelis, a leading provider of intelligent, high-performance networking solutions, announced the successful qualification and integration of its CN5000 Omni-Path® networking solution across Lenovo’s ThinkSystem V3 and V4 servers. The CN5000 is now a fully supported networking fabric in Lenovo’s EveryScale Solution, empowering customers to deploy tightly integrated, high-performance AI/HPC infrastructure at scale.

This integration enables Lenovo customers, especially those deploying artificial intelligence (AI), high performance computing (HPC) and technical computing workloads, to eliminate the network bottlenecks that commonly limit the efficiency of accelerated compute. The Cornelis CN5000 delivers ultra-low latency powered by a lossless and congestion free network, ensuring predictable scaling as AI/HPC clusters grow.

The Cornelis CN5000 integration on Lenovo platforms includes:

• Lenovo ThinkSystem V4: Featuring Intel® Xeon® 6 Processors
• Lenovo ThinkSystem V3: Featuring 4th Generation AMD EPYC™ 9004 Processor Series and the 4^th and 5^th Gen Intel® Xeon® Scalable Processors
• Air and liquid-cooled server variants for dense AI/HPC deployments including accelerated systems

“AI and HPC workloads are evolving rapidly—from single-node training to massive, distributed clusters that demand seamless inter-node communication,” said Scott Tease, Vice President, General Manager AI and HPC, Infrastructure Solutions Group, Lenovo. "By integrating end-to-end networking solutions like the Cornelis CN5000 with our ThinkSystem servers, we are giving customers the ability to securely scale their AI/HPC infrastructure with confidence—removing communication bottlenecks, improving system utilization, and accelerating time-to-insight. The result is lossless fabric, ultra-low latency, scalability, and interoperability—powering HPC speedups in CFD, climate modeling, and seismic simulations, while accelerating AI for science in material science, chemistry, weather forecasting, and astrophysics—all with energy-efficient HPC clusters.”

The CN5000 end-to-end 400 Gbps portfolio—comprising an ultra-low latency PCIe 5.0 SuperNIC and a 48-port, full non-blocking edge switch—has been purpose-built to meet the stringent inter-node communication demands of large-scale accelerated computing workloads. Uniquely aligned with emerging Ultra Ethernet requirements, the CN5000 delivers a fully lossless, congestion-aware transport with adaptive routing, per-flow credit-based flow control, and in-network telemetry. Its open-source software stack and field proven architecture ensure flexibility and interoperability across diverse AI/HPC environments. By eliminating communication bottlenecks in distributed AI training and HPC simulation workflows, the CN5000 allows organizations to fully realize the performance and efficiency potential of their Lenovo ThinkSystem servers and EveryScale solutions.

“The integration of Cornelis CN5000 into Lenovo’s EveryScale portfolio marks a pivotal moment in the evolution of scalable AI and HPC infrastructure,” said Lisa Spelman, CEO of Cornelis. “This is more than solving a network bottleneck. It’s about enabling organizations to finally match their compute ambitions with a fabric built for performance at scale. Lenovo’s leadership in delivering open, energy-efficient platforms combined with the CN5000’s ultra-low latency and lossless architecture unlock the full promise of accelerated computing. Together, we are redefining what’s possible in high-performance infrastructure.”

The qualification effort involved extensive collaboration between Cornelis and Lenovo delivering:

• Lenovo EveryScale 25B 2.1 Best Recipe: With Cornelis air-cooled CN5000 SuperNIC, switch and cables, and storage solutions
• Lenovo ThinkSystem V3 & V4 integration: Fully validated options including GPU accelerated systems with both AMD and Intel CPUs