Arista Networks has officially launched its 7060XE7 Series, a groundbreaking portfolio of 1.6T networking platforms engineered specifically for the demanding requirements of AI infrastructure. This new family represents a significant strategic shift for the company, moving from standalone high-performance switches to integrated rack-scale systems capable of handling the extreme density, power, and thermal efficiency that modern AI workloads demand.

Architecture and Silicon

The 7060XE7 Series is built on Broadcom's Tomahawk 6 silicon, which provides the 1.6T throughput needed for scale-up and scale-out AI fabrics. Arista is also collaborating with AMD on next-generation compute silicon and NICs, aiming to enable seamless end-to-end AI networking. This partnership is critical for delivering low-latency, high-bandwidth connections between GPUs and XPUs in large clusters.

The portfolio includes several distinct models tailored to different deployment scenarios. The 7060XE7-64PS and 7060XE7-64PRS are 4U air-cooled rack switches supporting pluggable Integrated heat sink (IHS) and Riding heat sink (RHS) optics. IHS is designed for current air-cooled data centers, while RHS targets future liquid-cooled AI fabrics requiring extreme port density. These models will be available in Q4 2026.

For high-density clusters, Arista offers the 7060XE7-64PRS-RV3-L, a specialized 2OU liquid-cooled platform featuring 224G SerDes. This system uses DC power from the ORv3 rack and contains no internal fans, integrating directly with liquid-cooled XPU servers to maximize power efficiency. Availability is expected in Q1 2027.

Additionally, the 7060XE7-128PE provides 128 800G ports in an air-cooled 4RU design with 100G SerDes, offering deployment flexibility and backward compatibility for environments not yet ready for full liquid cooling. This model also ships in Q1 2027.

Software and Open Standards

All 7060XE7 platforms run Arista's Extensible Operating System (EOS), which includes advanced low-latency and intelligent packet buffering to manage the intense microbursts characteristic of AI communication and collective patterns. EOS is complemented by support for open-source software such as Software for Open Networking in the Cloud (SONIC) and OpenSwitch, giving customers flexibility in network operating system choice.

A key software feature is full support for the Open Compute Project's Multipath Reliable Connection (MRC) protocol. MRC is an RDMA-based transport protocol that enables a single reliable connection to simultaneously utilize many network paths over Ethernet. According to Arista's leadership, MRC allows endstation NICs to stripe traffic across multiple links and paths, automatically handling out-of-order packets and responding to congestion signals like ECN and packet trimming. This results in very high fabric utilization with excellent load balancing, while interoperating with scale-across and WAN networks using standard dynamic routing protocols.

The software stack also includes robust load balancing, congestion management, telemetry, and diagnostics—all essential for optimizing AI networking performance. These capabilities allow administrators to monitor network health in real time and dynamically adjust traffic flows to avoid hot spots and link failures.

Ecosystem and Market Context

Arista's new family joins a rapidly expanding ecosystem of vendors targeting the 1.6T Ethernet market, including Cisco, Nvidia, Celestica, and others. The shift toward higher bandwidth, better power efficiency, and tighter integration between compute, optics, silicon, cooling, and network operating software is a defining trend in AI infrastructure. Industry analysts have highlighted the strong customer and ecosystem validation Arista has received from major cloud providers such as Microsoft Azure, Oracle Cloud Infrastructure, and Meta, as well as silicon partners AMD and Broadcom.

The 7060XE7 Series arrives at a time when AI workloads are placing unprecedented demands on data center networks. Traditional Ethernet designs often struggle with the bursty, loss-sensitive traffic generated by distributed training and inference. Arista's approach—combining purpose-built silicon, advanced cooling, and intelligent software—aims to address these challenges head-on, providing a foundation for next-generation AI fabrics.

As AI models grow larger and more complex, the need for reliable, high-throughput, low-latency networking becomes even more critical. Arista's rack-scale systems are designed to scale linearly with cluster size, supporting thousands of endpoints in a single fabric. The use of liquid cooling in certain models also reflects the industry's move toward more efficient thermal management to cope with the heat generated by densely packed switching silicon and optical transceivers.

For enterprises planning AI infrastructure investments, the 7060XE7 Series offers a clear migration path from traditional air-cooled deployments to future liquid-cooled environments. The support for both IHS and RHS optics means customers can start with current technology and upgrade to higher-density optics without replacing the entire switch. This forward-compatibility is a key differentiator in a market where technology cycles are accelerating.

Furthermore, Arista's commitment to open standards and interoperability ensures that the 7060XE7 Series can integrate into existing multi-vendor environments. The support for MRC, in particular, provides a standards-based way to achieve the performance and reliability that AI apps require, without locking customers into proprietary protocols.

From a deployment perspective, the rack-scale design reduces cabling complexity and improves power distribution. By integrating power, cooling, and networking into a single chassis, Arista simplifies the physical layout of AI clusters and reduces the number of individual components that must be managed. This holistic approach is expected to lower total cost of ownership over the lifecycle of the infrastructure.

Looking ahead, Arista is likely to continue evolving its AI networking portfolio as higher-speed Ethernet standards emerge. The collaboration with AMD on next-generation compute and NICs suggests a roadmap that will push beyond 1.6T in the coming years, potentially targeting 3.2T or even 6.4T fabrics. For now, the 7060XE7 Series positions Arista as a serious contender in the race to build the networking backbone for the AI era.

Source: Network World News