Ultra Ethernet Consortium Launches UEC 1.0 Specification for AI Networks and High-Performance Computing

The Ultra Ethernet Consortium (UEC) has officially announced the release of the UEC 1.0 specification, a communication stack based on Ethernet specifically designed to meet the demands of modern workloads in artificial intelligence (AI) and high-performance computing (HPC). This advancement marks a key milestone in redefining Ethernet as the foundation for next-generation data-intensive infrastructures.

A New Era for Ethernet in AI and HPC

The UEC 1.0 specification provides a high-performance, scalable, and interoperable solution across all layers of the networking stack, encompassing network interface cards (NICs), switches, optics, and cables. This allows for seamless integration between different vendors and accelerates innovation throughout the technology ecosystem.

One of the fundamental pillars of UEC is the promotion of open and interoperable standards, which prevent vendor lock-in and encourage widespread adoption within the industry. There are already active implementations and compliance programs in development, facilitating the deployment of UEC in data centers, cloud infrastructures, and enterprise environments.

Key Features of UEC 1.0

  • Modern RDMA over Ethernet and IP: UEC integrates remote direct memory access (RDMA) technologies, optimized for low-latency and high-performance environments, essential for AI and HPC tasks.
  • Open Standards and Compatibility: The focus on interoperability eases technology adoption without relying on a single vendor.
  • End-to-End Scalability: UEC is designed to scale from basic network operations to managing and testing millions of nodes, adapting to the needs of large cloud operators and next-generation data centers.

Easy Integration and Real Benefits

Based on the widely adopted Ethernet standard globally, UEC 1.0 simplifies deployment across the technology stack, from hardware to applications. This familiarity, along with robust tools, makes migration and adoption by DevOps teams, cloud infrastructure managers, hyperscalers, and technical leaders in AI straightforward and low-risk.

“The Ultra Ethernet 1.0 standard is the result of exceptional collaboration among experts in AI, HPC, networking, system and silicon manufacturers, and network operators. It includes advancements in transport protocols, congestion control, DMA, Ethernet link technologies, and network security,” emphasized Hugh Holbrook, Chair of the UEC Technical Advisory Committee.

A Milestone for the Tech Ecosystem

The publication of the UEC 1.0 specification marks the beginning of a new phase for connectivity in high-demand computational environments. The consortium underscores that this launch is just the start of a more ambitious roadmap, with Ultra Ethernet set to lead the scalability and efficiency of networks that will drive artificial intelligence and high-performance computing in the coming years.

As Dr. J Metz, Chair of the Consortium’s Steering Committee, noted: “This standard is the culmination of unprecedented collaboration in the industry and provides the intelligent, low-latency, high-bandwidth transport required for today’s and tomorrow’s most demanding workloads. Ultra Ethernet is here, and it’s built to scale the future.”

Access the Ultra Ethernet 1.0 specification.

Scroll to Top