As AI and high-performance computing (HPC) clusters grow, power, cooling, and network congestion have become the primary bottlenecks to scalability. Cornelis Networks and Supermicro are addressing these intertwined challenges through a new validated infrastructure solution. By integrating Cornelis's high-performance CN5000 networking with Supermicro's liquid-cooled FlexTwin server platforms, the collaboration aims to enable larger, more efficient AI/HPC deployments within constrained data center environments.
Cornelis and Supermicro validate a joint solution for AI/HPC infrastructure bottlenecks.
It combines Supermicro's liquid-cooled FlexTwin servers with Cornelis's CN5000 400Gbps networking.
The solution targets constraints in power, cooling, and network congestion that limit cluster scaling.
Joint testing showed up to 1.5x higher application performance vs. other 400G fabrics.
It also delivered up to 2.3x higher performance per watt in liquid-cooled configurations.
The integration allows data centers to run larger workloads without increasing power draw.
The performance of large-scale AI training and scientific simulation is no longer limited solely by processor speed. Network communication delays and thermal/power constraints now dictate real-world scalability and workload completion times. This collaboration directly targets these "gating factors" by providing a cohesive hardware stack where advanced networking and efficient cooling are designed to work in concert, preventing one component from bottlenecking the entire system.
The solution brings together two specialized technologies. Cornelis's CN5000 networking platform provides predictable, congestion-free 400Gbps data movement between servers, which is critical for parallel processing workloads. Supermicro's FlexTwin platform employs a dense, liquid-cooled design that captures up to 95% of system heat, enabling stable operation within strict power and thermal limits. The CN5000's support for both air-cooled and liquid-cooled configurations ensures compatibility across various data center environments.
Joint testing across manufacturing, life sciences, and climate workloads demonstrated significant advantages. The combined solution delivered up to 1.5 times higher application performance compared to other 400G network fabrics and an average 29% improvement in performance per networking dollar. Most critically for power-constrained facilities, it achieved up to 2.3 times higher performance per watt. These gains translate to the ability to solve larger problems faster, use fewer nodes, and stay within fixed power budgets.
This validated partnership provides a practical blueprint for enterprises and research institutions hitting the walls of power and network scalability. By treating the compute, network, and thermal design as an integrated system, Cornelis and Supermicro offer a path to sustain the exponential growth of AI and HPC workloads. It represents a shift in infrastructure strategy, where efficiency and interconnect performance are as critical as raw compute power for achieving next-generation results.
About Cornelis
Cornelis Networks delivers high-performance, scale-out networking solutions that accelerate AI and HPC workloads. Built on the powerful Omni-Path architecture, Cornelis technology enables lossless, congestion-free networking that reduces training time, improves inference, and maximizes compute utilization. From foundation model training to complex climate modeling and real-time analytics, Cornelis' solutions power the most demanding workloads across commercial, academic, government and cloud environments. With a focus on performance, scalability, and efficiency, Cornelis helps organizations achieve faster insights and greater return on infrastructure investments.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.