Advancing open Ethernet for AI: HPE Networking joins the ESUN initiative

[NetworkExperts]

HPE offers scalable, open Ethernet solutions for AI clusters and data centers. Discover how HPE and ESUN create efficient, interoperable AI data center networks.

The AI era is redefining what networks need to deliver. The massive compute scale of today’s AI systems demands unprecedented performance, lossless communication, and efficient data movement—whether across nodes, racks, or clusters.

That’s why HPE is proud to be part of the Ethernet for Scale-Up Networking (ESUN) initiative, a new industry collaboration under the Open Compute Project (OCP). The goal of ESUN is clear: to evolve open Ethernet standards to meet the needs of scale-up AI infrastructure—interconnecting accelerators/XPUs within clusters—with the same openness, interoperability, and performance that Ethernet has long delivered for scale-out systems.

Open networking for AI

At HPE, we believe AI infrastructure should be open and accessible—not locked behind proprietary fabrics. ESUN represents a meaningful step toward that vision. By defining an open Ethernet standard for scale-up interconnects, the industry can help ensure that any XPU or GPU can communicate efficiently over a common, standards-based network.

Scale-up networking introduces new requirements: Ultra-low latency, deterministic packet delivery, congestion management, and adaptive load balancing. These are challenges we’ve already been tackling at HPE as part of our AI networking platform.

Proven leadership

Our experience in scale-out networking has set the foundation for what comes next:

• High performance AI fabrics: With the Juniper QFX Series Switches, we’ve delivered some of the industry’s most performant and scalable Ethernet fabrics for AI clusters. HPE was the first pure-play networking vendor to publish MLPerf benchmarks under MLCommons — demonstrating the power of open Ethernet for AI workloads. We’ve introduced AI load balancing: an innovation designed to virtually mitigate congestion and boost utilization in Ethernet-based AI networks.
• AI networking: Our AI networking platform delivers both networking for AI and AI for networking. With breakthrough technology from Apstra and Juniper Mist, we’re bringing AIOps-driven automation to design, deploy, and operate complex AI clusters, delivering real-time insights, closed-loop assurance, and faster operational velocity.
• Scale-across strength: Juniper PTX Series Routers, with 800G capabilities, provide the backbone for large-scale, high-throughput connectivity across data center regions and clouds.

Together, these innovations form the backbone of open, high-performance networking for AI at every scale—across nodes, clusters, and data centers.

HPE’s leadership continues to earn industry-wide recognition. Juniper Networks (now HPE) was named a leader in 2025 Gartner® Magic Quadrant™ for Data Center Switching, and we secured #2 position in AI Ethernet Fabric Build-out use case in Gartner’s Critical Capabilities report—a testament to our relentless innovation in delivering open, automated, and truly AI networking solutions

Looking ahead: Scale up, scale across, and scale out, together

We view ESUN as a critical bridge—extending Ethernet from scale-out and scale-across to scale-up. It’s a natural evolution that ensures the AI ecosystem can grow without fragmenting proprietary solutions.

As the industry advances through ESUN, HPE shall continue to contribute expertise in congestion management, performance telemetry, and open automation, helping shape a unified Ethernet framework for AI networking.

For more information on HPE data center solutions, visit the following links.

Juniper AI data center solutions

Introducing ESUN: Advancing Ethernet for scale-up AI infrastructure at OCP

Meet the author:

Praful Lalchandani
VP of Product, DC Platforms and AI Solutions, HPE