Arista Unveils 7060XE7, a 1.6-Terabit Switch Family Built on Broadcom's Tomahawk 6, Targeting Rack-Scale AI Fabrics

Overview

Arista Networks on June 9 introduced the 7060XE7 Series, a portfolio of 1.6-terabit Ethernet switches aimed at the back-end networks that stitch together AI accelerators, according to Arista Networks. The new platforms are built on Broadcom’s Tomahawk 6 silicon, the 102-terabit-per-second switch chip that has become the common foundation for this generation of high-bandwidth Ethernet systems. Arista describes the series as an expansion of its Etherlink family for both scale-out and scale-up applications, per the company’s announcement.

The launch places Arista alongside a growing group of vendors pushing 1.6T Ethernet into AI clusters that are expanding from thousands to hundreds of thousands of accelerators, as reported by SDxCentral.

What We Know

The 7060XE7 portfolio spans four models with different cooling and port configurations, according to Arista:

• The 7060XE7-64PS and 7060XE7-64PRS are air-cooled, 4RU systems with 64 ports of 1.6T, slated for availability in the fourth quarter of 2026.
• The 7060XE7-64PRS-RV3-L is a liquid-cooled, 2RU switch, also with 64 ports of 1.6T, using 224G SerDes and carrying no internal fans, with availability set for the first quarter of 2027.
• The 7060XE7-128PE is an air-cooled, 4RU system with 128 ports of 800G built on 100G SerDes, likewise due in the first quarter of 2027.

All four models run Arista’s Extensible Operating System (EOS). Arista’s software set for the platforms includes Dynamic Load Balancing, Cluster Load Balancing, and Multipath Reliable Connection (MRC) for steering traffic around congestion, along with Link Layer Retry and congestion-signaling features, per the Arista announcement. Network World reports that the switches offer full Open Compute Project support for MRC, which shifts load to better-performing paths in response to network congestion.

Tyson Lamoreaux, Senior Vice President, Cloud and AI Networking at Arista Networks, framed the launch around a changing role for the network. “The AI era requires a shift in how we think about the network - it is no longer a standalone layer of infrastructure, but a tightly integrated component of the AI supersystem,” he said in the company statement. “With the 7060XE7 Series, we are delivering massive-scale 1.6T systems that combine world-class reliability and the differentiation of EOS with liquid cooling and low-power optics to help our customers build AI fabrics designed for maximum performance and power efficiency.”

The announcement drew supporting statements from several of Arista’s largest customers and silicon partners, all quoted in the Arista press release. Rani Borkar, President, Azure Hardware Systems and Infrastructure at Microsoft, tied the 1.6T interface to the company’s own AI hardware: “Our collaboration with Arista on the 1.6T Ethernet interface helps enable the next generation of AI clusters with greater interconnect capacity for Azure Maia, Microsoft’s AI accelerator, and Fairwater, Microsoft’s extreme-scale AI datacenters, while preserving operational simplicity across our infrastructure.”

Gaya Nagarajan, Vice President of Infrastructure at Meta, said that “Arista’s 1.6T platforms and liquid-cooled designs align with our focus on open, scalable AI fabrics that meet the requirements of next-generation training and inference,” while Mahesh Thiagarajan, Executive Vice President, Oracle Cloud Infrastructure, said the platforms “provide the throughput, determinism, and stability needed for our RDMA-based AI fabrics.” All three quotes appear in the Arista announcement.

On the silicon side, Dr. Asad Khamisy, Senior Vice President and General Manager, Core Switching Group at Broadcom, said the collaboration works “By powering Arista’s 1.6T Ethernet solutions with Broadcom’s industry-leading Tomahawk 6 102-Tbps Ethernet switches,” according to the Arista press release. Network World reports that Arista is also working with AMD on next-generation compute silicon and network interface cards to support scale-out AI fabrics.

What We Don’t Know

Arista has not published pricing for the 7060XE7 line, and the two liquid-cooled and 800G configurations are not scheduled to ship until early 2027, leaving real-world performance and adoption figures unavailable at launch. The named hyperscalers voiced support in the announcement, but neither Arista nor its partners disclosed deployment volumes or firm purchase commitments.

Analysis

The 7060XE7 arrives as Ethernet consolidates its position in AI back-end networks. The Tomahawk 6 foundation is now shared across multiple switch makers, meaning much of the competition is shifting to software, cooling, and system integration rather than raw silicon bandwidth. SDxCentral notes that Ethernet is expanding into scale-up networking, the domain long dominated by Nvidia’s NVLink, a contest that has also drawn newer entrants such as the UALink camp, as previously reported.

The launch also underscores how crowded the 1.6T Ethernet field has become. Broadcom’s Tomahawk 6 competes with Cisco’s own Silicon One G300, while Nvidia’s Spectrum-X recently overtook the field to become the top data center Ethernet switch vendor by revenue. Arista’s bet is that its EOS software stack, liquid-cooled hardware, and long-standing hyperscaler relationships will keep it competitive as those customers scale their AI fabrics.