AIToday
Open-Source AITop Companies' AI MovesAI Business & IndustryTop Companies AIPublished: Oct 12, 2026, 06:30 JST

Arista Networks unveils open Ethernet rack for 144 accelerators

Arista Networks unveils open Ethernet rack for 144 accelerators

3 Key Points

  1. What happened

    Arista Networks announced an open Ethernet rack-scale portfolio, engineered with AMD, Arm, Broadcom, d-Matrix, Meta, Microsoft, and Qualcomm, supporting up to 144 accelerators per rack and more via cross-rack setups.

  2. Why it matters

    It pitches an open, standards-based alternative to proprietary compute islands, letting cloud operators scale AI infrastructure with their preferred architecture and, per Arista, optimize token economics without lock-in.

WHO IT HITSCloud titan operators and enterprise AI infrastructure teams designing high-density AI clusters gain a standards-based Ethernet option, with Arista and its compute partners positioning it as an alternative to proprietary interconnects.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

The announcement positions Arista's Ethernet-based rack designs against the PCIe and proprietary interconnects that have constrained tightly coupled communication across large accelerator pools as AI models expand. By partnering with AMD, Arm, Broadcom, d-Matrix, Meta, Microsoft, and Qualcomm, Arista is aligning its hardware with a broad set of compute vendors rather than a single chip ecosystem.

The portfolio spans several physical form factors — an orthogonal chassis for single-rack density and a cabled backplane for serviceability — giving operators options depending on accelerator and workload needs. Arista frames the open, standards-based approach as a way for cloud titans to scale heterogeneous compute without proprietary lock-in.

FAQ
How many accelerators can Arista's rack-scale architecture support?
Arista says the designs support up to 144 accelerators per rack. Its Etherlink SU-144 scales to 1,024 XPUs through cross-rack topologies.
What cooling does the new architecture use?
The orthogonal chassis design offers 100–400kW liquid cooling support. Arista describes the racks as liquid-cooled and built for extreme density, power, and thermal demands.
What is ESUN?
ESUN stands for the Ethernet for Scale-up Networks initiative, which Arista says this effort leverages. It underpins Arista's open Ethernet approach to scale-up domains.
Top Companies AIRead Original Article

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleELYZA RSI Research targets AI developing AI