Arista has introduced the Arista 7060XE7 Series, a 1.6T switching portfolio aimed at rack-scale AI fabrics where the network is treated as part of a tightly integrated “AI supersystem,” not a standalone tier. The company is pitching the platforms at the extreme density, power, and thermal constraints that show up as AI clusters scale to very large XPU counts.
The 7060XE7 line is positioned as a mix of fixed switch platforms and configurable rack-scale systems, with designs intended to fit air-cooled, liquid-cooled, and hybrid-cooled deployments. Arista also calls out low-latency behavior and “intelligent packet buffering” targeted at AI traffic patterns with heavy microbursts, plus a suite of Arista EOS features it says AI fabrics typically require.
For data center engineers, the practical takeaway is that networking gear is increasingly being specified alongside rack power delivery and cooling topology, not after the fact. A switch that’s explicitly designed around ORv3 racks, fanless operation, and optics power targets can change how you think about rack integration, serviceability, and heat rejection paths—especially as fabrics push higher port rates and higher aggregate throughput.
Platforms and configurations
Arista lists several 7060XE7 configurations:
The 7060XE7-64PS and 7060XE7-64PRS are 4RU air-cooled rack switches that support both Integrated heat sink (IHS) and Riding heat sink (RHS) optics.
The 7060XE7-64PRS-RV3-L is a 2OU liquid-cooled platform intended for high-density clusters. It uses 224G SerDes, draws DC power from the ORv3 rack, and has no internal fans, with the goal of integrating alongside liquid-cooled XPU servers.
The 7060XE7-128PE provides 128 800G ports in a 4RU air-cooled design using 100G SerDes, targeting deployment flexibility and backward compatibility.
Throughput, optics, and silicon ecosystem
Arista says the 7060XE7 Series uses the latest SerDes technology and provides 100 Terabits/sec of bandwidth, with “1.6 terabits /sec throughput across ports.” The company also highlights support for Linear Pluggable Optics (LPO), claiming roughly a 60% reduction in interconnect power consumption.
On partnerships, Arista says it is working with AMD on next-generation compute silicon and NICs for scale-out AI fabrics, and it highlighted Broadcom Tomahawk 6 silicon for its 1.6T Ethernet solutions. “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,” said Tyson Lamoreaux, senior vice president, Cloud and AI Networking at Arista.
Availability
Arista lists availability as Q4 ’26 for the 7060XE7-64PS/PRS (64x 1.6T, air-cooled), Q1 ’27 for the 7060XE7-64PRS-RV3-L (64x 1.6T, liquid-cooled), and Q1 ’27 for the 7060XE7-128PE (128x 800G, air-cooled).
Source: Arista
