New CN5000 integration with Supermicro FlexTwin™ server platforms allows data centers to run larger AI and HPC simulation workloads using less power and infrastructure

As enterprise data centers struggle to run increasingly complex AI training and scientific simulation workloads within tight power and cooling limits, Cornelis and Supermicro have validated a new high-performance infrastructure solution designed to deliver more computing power using less energy. Supermicro’s liquid-cooled FlexTwin™ server platforms are now fully validated with Cornelis’ CN5000 high-speed networking, giving organizations a practical way to scale AI and HPC clusters without hitting performance, power, or cost walls.

“We are excited that Supermicro’s FlexTwin platforms are fully validated for the Cornelis CN5000 networking option,” said Vik Malyala, President & Managing Director EMEA, SVP Technology & AI, Supermicro. “By integrating the Cornelis CN5000 networking, we’re giving our customers the tools they need to accelerate the most demanding HPC and AI workloads without compromise.”

The Infrastructure Challenge

AI model training, advanced simulation, and data-intensive research increasingly depend on large clusters of connected processors and accelerators that operate in parallel. While compute performance continues to rise, power availability, cooling capacity, and network congestion have become the dominant constraints on real-world performance. In many environments, the network, not the processors, now determines workload completion speed and system scalability.

“The infrastructure demands of AI and modern simulation have fundamentally changed, with power, cooling, and network scalability emerging as the new gating factors,” said Lisa Spelman, CEO of Cornelis Networks. “By combining Supermicro’s advanced liquid-cooled FlexTwin architecture with our high-performance CN5000 networking platform, customers gain a proven way to scale performance without paying the traditional penalties in power, cost, or complexity.”

A Practical Path to Higher Performance with Lower Energy Use

Cornelis’ CN5000 400Gbps networking platform addresses the communication bottleneck by providing predictable, congestion-free data movement between servers—one of the most critical and most overlooked factors in large AI and HPC deployments. Supermicro’s FlexTwin platform addresses the thermal side of the equation with a dense, liquid-cooled design capable of capturing up to 95% of system heat, enabling stable performance in power-constrained data centers. In addition, the CN5000 portfolio supports both air-cooled and liquid-cooled configurations across its switches, as well as air-cooled and conduction-cooled options for the CN5000 SuperNIC—giving Supermicro FlexTwin customers the flexibility to deploy high-performance networking in any thermal environment.

Together, the validated solution enables organizations to:

• Run larger, faster AI training and HPC simulations.
• Increase application throughput without increasing power draw.
• Reduce total cost of ownership through higher efficiency per rack.
• Maintain stable performance as clusters scale in size.

Proven Performance and Efficiency Gains

In joint testing across manufacturing, physics, life sciences, and climate workloads, the Supermicro FlexTwin + Cornelis CN5000 solution delivered:

• Up to 1.5× higher application performance versus competitive 400 Gbps network fabrics.
• An average 29% improvement in performance per networking dollar.
• Up to 2.3× higher performance per watt in liquid-cooled configurations.
• Improved scaling efficiency, allowing clusters to grow without the typical performance drop-off.

These gains allow data center operators to solve larger problems faster, deploy fewer total nodes for the same workload, and operate within fixed power budgets.

The Supermicro FlexTwin platform with Cornelis CN5000 networking is available now through Supermicro and authorized Cornelis partners. For more information and details on the joint testing and benchmarked performance and cost advantages, please read the Solution Brief.

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. Learn more at cornelis.com.

Cornelis and Supermicro AI & HPC Infrastructure Solution FAQ

1. What is Cornelis’ CN5000 and what is Supermicro’s FlexTwin? Cornelis’ CN5000 is a high-speed 400 Gbps networking platform designed to move data between servers with low latency and no congestion—critical for large AI and HPC clusters. Supermicro’s FlexTwin is a dense, liquid-cooled server platform engineered to run high-performance workloads efficiently within strict power and cooling limits and is designed to house the most performant and latest generation of CPUs from Intel and AMD. Cornelis CN5000 and Supermicro FlexTwin together form an integrated solution for scaling AI and scientific computing.
   
   
   
   
2. What’s driving power and network bottlenecks in modern AI and HPC data centers? As AI models grow and simulations become more complex, thousands of GPUs and CPUs must exchange massive volumes of data in real time. While compute performance continues to increase, power availability, cooling capacity, and network congestion now limit how fast workloads can run and how large clusters can scale, making infrastructure efficiency a top operational constraint.
   
   
   
   
3. How does the CN5000 integration with FlexTwin address these bottlenecks, and what results can customers expect? FlexTwin’s liquid cooling removes heat efficiently, enabling high-density compute to operate at full speed, while CN5000 provides predictable, congestion-free data movement between servers. In joint testing, the integrated solution delivered up to 1.5× higher performance, 29% better performance per networking dollar, and up to 2.3× higher performance per watt in liquid-cooled environments.

View source version on businesswire.com: https://www.businesswire.com/news/home/20251209720521/en/