Arista Networks (NYSE: ANET) announced a rack-scale open Ethernet portfolio for AI infrastructure, developed in partnership with AMD, Arm, Broadcom, d-Matrix, Meta, Microsoft, and Qualcomm. The portfolio targets both scale-up and scale-out AI networking deployments and is based on the Ethernet for Scale-up Networks (ESUN) initiative.

The scale-up offering, called Etherlink SU-144, supports up to 144 accelerators in a single-hop network and scales to 1,024 accelerators through cross-rack configurations. Arista is offering three physical form factors: an orthogonal chassis supporting 100–400kW liquid cooling; a cabled backplane with a modular rack structure; and a cross-rack system for multi-rack expansion.

The scale-out networking solution is built on Arista's Netdi platform and supports multiple scale-up protocols, including ESUN, Nvidia's NVLink, NVLink Fusion, and PCIe. It incorporates Ultra Ethernet Consortium standards, Multipath Reliable Connection fabric resiliency, and Dynamic Load Balancing. Each rack includes multiple liquid-cooled 7060EX7 switches with 102.4 Tbps capacity per switch.

Arista states that next-generation optics, including XPO and open CPO/NPO formats, will increase rack capacity from 1.6 Pb/s with current OSFP-based switches to 6.5 Pb/s, which the company claims translates to nearly a 50% reduction in data center footprint.

"The network in the AI era is no longer just an independent pipe between servers. It's now a tightly integrated rack-scale backplane that is essential to maximizing compute," said Tyson Lamoreaux, Senior Vice President, Cloud and AI Networking at Arista Networks.

Arista plans to demonstrate the rack-scale networking solutions at the OCP Global Summit, scheduled for October 12–15, 2026.