HomeTechnologyArtificial IntelligenceIntroducing Ethernet Scale-Up Networking: Advancing Ethernet for Scale-Up AI Infrastructure

Introducing Ethernet Scale-Up Networking: Advancing Ethernet for Scale-Up AI Infrastructure

Courtesy: Broadcom

With the increase in scale and complexity of AI systems, Ethernet is once again evolving to meet the challenge. At the OCP Global Summit 2025, Broadcom, along with AMD, ARM, Arista, Cisco, HPE Networking, Marvell, Meta, Microsoft, NVIDIA, OpenAI, and Oracle, announced a new collaborative effort called Ethernet for Scale-Up Networking (ESUN).

With the initiation of this workstream, the OCP Community now has the opportunity to address areas that enhance scale-up connectivity across accelerated AI infrastructure. The scale-up domain in XPU-based systems can be viewed in two primary areas: 1) network functionality, and 2) XPU-endpoint functionality.

First, the network aspect of scale-up focuses on how traffic is sent out across the network switches themselves, including protocol headers, error handling, and lossless data transfer.  This is what ESUN intends to address and what the planned OCP workstream by the same name will focus on. The workstream itself is planned to kick off shortly after the OCP Global Summit.

Second, in the XPU-endpoint domain, the design depends on factors such as workload partitioning, memory ordering, and load balancing, and it is often tightly co-designed with the XPU architecture itself.

What is ESUN?

ESUN is a new workstream collaboration designed as an open technical forum to advance Ethernet in the rapidly growing scale-up domain for AI systems. This initiative brings together operators and leading vendors to collaborate on leveraging and adapting Ethernet for the unique demands of scale-up networking.

Key Focus Areas:

  • Technical Collaboration: ESUN serves as an open forum where operators, equipment and component manufacturers can jointly advance Ethernet solutions optimized for scale-up networking.
  • Interoperability: The initiative emphasizes the development and interoperability of XPU network interfaces and Ethernet switch ASICs for scale-up.
  • Technical Focus: Initial focus will be on L2/L3 Ethernet framing and switching, enabling robust, lossless, and error-resilient single-hop and multi-hop topologies.
  • Standards Alignment: ESUN will actively engage with organizations such as UEC (Ultra-Ethernet Consortium) and IEEE 802.3 (Ethernet) to align with open standards, incorporate best practices, and accelerate innovation.
  • Ecosystem Enablement: By leveraging Ethernet’s mature hardware and software ecosystem, ESUN will encourage diverse implementations and drive rapid adoption across the industry.
Ethernet for Scale-Up Networking

What Are the Focus Areas for ESUN?

ESUN focuses solely on open, standards-based Ethernet switching and framing for scale-up networking—excluding host-side stacks, non-Ethernet protocols, application-layer solutions, and proprietary technologies.

How is this Different from SUE?

OCP has previously launched an effort to advance the endpoint functionality for scale-up networking through the SUE-Transport (Scale-Up Ethernet Transport) workstream (originally named SUE; it has been renamed and clarified as SUE-T in contrast to ESUN). SUE-T will carry forward some of the SUE work, which was seeded with the Broadcom contribution of the version 1.0 specification.

ELE Times Research Desk
ELE Times Research Deskhttps://www.eletimes.ai
ELE Times provides extensive global coverage of Electronics, Technology and the Market. In addition to providing in-depth articles, ELE Times attracts the industry’s largest, qualified and highly engaged audiences, who appreciate our timely, relevant content and popular formats. ELE Times helps you build experience, drive traffic, communicate your contributions to the right audience, generate leads and market your products favourably.

Related News

Must Read

Top 10 Advanced IC Substrate Manufacturers in 2026: Leading Companies Powering the AI Semiconductor Revolution

A combination of Artificial Intelligence (AI), High-Performance Computing (HPC),...

Cyient DLM and Hindustan Shipyard Sign MoU to Advance Indigenous Maritime and Defence Electronics

Partnership to Focus on Indigenous Electronics, Advanced Manufacturing and...

Driving Motion Precision: How Physical AI is Converting Data into Action

Autonomous manufacturing is entering a new phase where robotics,...

GÖPEL Electronic Joins Automotive SerDes Alliance to Support ASA-ML Automotive Networking

High-speed communication technologies are becoming increasingly important in modern...

Texas Instruments Uses Agentic AI to Cut MCU Software Migration from Months to Days

Texas Instruments (TI) has upgraded its CCStudio development ecosystem...

Microchip Unveils ATFCD52 for Real-Time Robotic Grip Control and Slip Detection

Reliable robot operation requires more than cameras, especially when...

Keysight Expands High-Power Test Portfolio With New 12 kW DC Power Supplies

Keysight Technologies has expanded its high-power automated test equipment...