Ultra Ethernet Stack | Open Networking for the Hyperscale and AI Era

As AI clusters grow from thousands to hundreds of thousands of accelerators, traditional networking approaches face increasing challenges related to scalability, congestion management, operational complexity, and interoperability. The Ultra Ethernet architecture addresses these challenges with a purpose-built transport layer, advanced congestion control, native security, and AI-optimized traffic management, while maintaining compatibility with standard Ethernet and IP infrastructure. 

Built by the Ultra Ethernet Consortium (UEC), Ultra Ethernet is designed to deliver the performance required for large-scale AI training, AI inference, and HPC environments without the limitations of proprietary networking fabrics. By combining open standards with advanced networking innovations, Ultra Ethernet enables organizations to build scalable, high-performance AI infrastructure using a multi-vendor ecosystem. 

This white paper provides an overview of the Ultra Ethernet stack and explores how next-generation AI data centers can benefit from improved network efficiency, simplified operations, and greater deployment flexibility. 
Ultra Ethernet Stack | Open Networking for the Hyperscale and AI Era
What You'll Learn 

  • Why AI and HPC workloads are driving the need for a new networking architecture beyond traditional Ethernet approaches.  
  • How Ultra Ethernet combines open, standards-based networking with the performance characteristics required for large-scale AI and HPC environments. 
  • The key innovations behind Ultra Ethernet Transport (UET), including out-of-order delivery, packet spraying, and IP-native operation. 
  • How advanced congestion control and native security help improve efficiency, scalability, and operational simplicity in AI infrastructure.  
  • How standardized traffic profiles optimize networking for AI training, inference, and HPC workloads on a shared fabric. 
  • How Ultra Ethernet is designed to support next-generation AI clusters with massive scale, high bandwidth, and low latency. 

Gain a practical introduction to the technologies shaping the future of AI networking and discover how open Ethernet-based architectures are evolving to support hyperscale AI and HPC deployments. 
Download the white paper to learn how the Ultra Ethernet Consortium (UEC) is redefining AI networking with an open, standards-based architecture engineered for performance, interoperability, scalability, and operational simplicity. 
By submitting this form, you are also granting VIAVI Solutions permission to send you emails about related topics in accordance with our Privacy Policy. You can opt-out at any time by clicking the unsubscribe link in our emails.