Arista Networks has taken the wraps off its 7060XE7 Series, a new portfolio of 1.6T networking platforms designed to provide the foundation for rack-scale AI infrastructure. This announcement, made on June 9, 2026, represents a significant step forward in the evolution of data center networking, as the industry shifts from traditional switch-centric designs toward holistic rack-scale systems that can handle the extreme demands of artificial intelligence workloads.

The 7060XE7 family features fixed switch platforms and configurable rack-scale systems, targeting racks for vertical and horizontal AI workflows. All will run Arista's Extensible Operating System (EOS), which includes low-latency and intelligent packet buffering to manage the intense microbursts typical of AI communication and collective patterns, Arista stated. This software foundation is critical because AI training jobs generate bursty, unpredictable traffic flows that can overwhelm conventional switch buffers.

Built on Broadcom Tomahawk 6 Silicon

The 7060XE7 family is built on Broadcom Tomahawk 6 silicon, which is the latest generation of the company's high-performance switching ASICs. Tomahawk 6 offers 51.2 Tbps of switching capacity and supports 1.6T Ethernet ports, enabling unprecedented throughput for AI clusters. Arista is also working with AMD on next-generation compute silicon and NICs to enable scale-out AI fabrics. The collaboration with AMD is strategic because AMD's MI300 and future accelerators require high-bandwidth, low-latency networking to scale efficiently across thousands of nodes.

The Tomahawk 6 chip employs advanced SerDes technology operating at 224G per lane, which allows for 1.6T ports using eight lanes each. This is a major leap from the previous generation (Tomahawk 5, which used 112G SerDes for 800G ports). The higher serial rate reduces the number of optical lanes needed, lowering power consumption and cost per bit.

Rack-Scale Systems for AI Workloads

Strategically, the 7060XE7 Series signifies Arista's transition from offering standalone, high-performance switches to providing rack-scale systems that can handle the extreme density, power, and thermal efficiency AI requires, Arista stated. The platforms allow customers to build scale-up and scale-out AI fabrics using air, liquid, and hybrid-cooled technology. This flexibility is essential because AI training clusters can consume tens of kilowatts per rack, and traditional air cooling may be insufficient for the highest-density configurations.

Specific Configurations

Arista detailed three specific models in the portfolio:

• 7060XE7-64PS and 7060XE7-64PRS 4U Rack Switches: Available in Q4, these air-cooled systems offer support for pluggable Integrated heat sink (IHS) and Riding heat sink (RHS) optics. IHS is aimed at current air-cooled data centers, and RHS would be aimed at future liquid-cooled AI fabrics and extreme port density, Arista stated. These switches provide 64 1.6T ports, delivering a massive 102.4 Tbps of aggregate switching capacity.
• 7060XE7-64PRS-RV3-L: This is a specialized 2OU liquid-cooled platform for high-density clusters, featuring 224G SerDes. This system uses DC power from the ORv3 rack and contains no internal fans, integrating with liquid-cooled XPU servers to maximize power efficiency. It will be available in Q1 2027. The elimination of fans reduces noise and improves reliability, as fans are often a major failure point in networking gear.
• 7060XE7-128PE: Also coming in Q1 2027, these devices provide 128 800G ports in an air-cooled 4RU design, utilizing 100G SerDes, for environments requiring deployment flexibility and backward compatibility. This model is designed for customers who need to connect existing 800G devices while preparing for a future migration to 1.6T.

Software Ecosystem and Open Networking

On the software side, EOS is the featured network operating system, but the family also supports open-source software such as Software for Open Networking in the Cloud (SONIC) and OpenSwitch. This dual support allows customers to choose between Arista's proprietary OS with advanced features like intelligent buffering and congestion management, or an open-source alternative that enables custom automation and integration with existing DevOps pipelines. The inclusion of SONiC is particularly notable because it is widely used in large-scale cloud data centers operated by Microsoft, Meta, and others.

Multipath Reliable Connection (MRC)

One of the portfolio's key features is the full support for the Open Compute Project's Multipath Reliable Connection (MRC). MRC is an RDMA‑based transport protocol that allows a single reliable connection to simultaneously use many network paths over Ethernet. This is a breakthrough for AI networking because traditional TCP or RDMA over Converged Ethernet (RoCE) often underutilize available bandwidth due to congestion and packet drops.

“MRC is an open protocol where endstation NICs stripe their traffic across multiple links and paths to the receiver, with out of order packets automatically handled,” wrote Arista's Kenneth Duda, president and CTO, and Alan Judge, distinguished engineer, in a blog about the technology. “MRC responds to network congestion signals (ECN and packet trimming), shifting load to the best-performing paths, and avoiding links and paths that can't actually reach the destination altogether.”

The protocol works by dividing data into subflows that traverse different physical paths. At the receiver, packets are reordered transparently. If a path becomes congested or fails, MRC quickly redistributes traffic to other healthy paths, ensuring high utilization and resilience. This is similar in spirit to how multipath TCP works, but optimized for RDMA and low-latency AI fabrics.

The software also supports load balancing, congestion management, telemetry and diagnostics, and other technologies that will be core to AI networking, Arista stated. For example, the switches can generate detailed telemetry data about queue depths, packet drops, and latency per flow, which can be fed into network analytics platforms for real-time optimization.

Market Context and Competition

The new Arista family joins a growing ecosystem of vendors looking to tap into the 1.6T Ethernet world, which includes Cisco, Nvidia, Celestica and others. Cisco has its own 1.6T switches based on the Silicon One architecture, while Nvidia is pushing its Spectrum-4 platform for AI. However, Arista's strength lies in its EOS operating system, which has a loyal customer base among large cloud providers and financial institutions. The 7060XE7 series also benefits from Arista's history of early adoption of new Broadcom ASICs, allowing the company to bring products to market quickly.

“Arista Network's new 7060XE7 Series is a strong signal of where large-scale AI fabrics are heading: higher bandwidth, better power efficiency, and tighter integration between compute, optics, silicon, cooling, and network operating software,” wrote Sameh Boujelbene, vice president, data center switch and AI networks market research for Dell Oro, in a LinkedIn post. Among the features that stand out to her are “strong customer and ecosystem validation from Microsoft Azure, Oracle Cloud Infrastructure, Meta, AMD, and Broadcom.”

Historical Context and Evolution

Arista was founded in 2004 by Andy Bechtolsheim, David Cheriton, and others who previously worked at Sun Microsystems and Cisco. The company pioneered the concept of a merchant silicon-based data center switch, using Broadcom chips to deliver high performance at lower cost than proprietary ASICs. Over the years, Arista has become a dominant player in the cloud networking space, competing head-to-head with Cisco. The 7060XE7 series continues this tradition by leveraging the latest merchant silicon to push the boundaries of speed and density.

The shift to rack-scale systems is not unique to Arista. Competitors like Cisco and Juniper are also offering integrated systems that combine switches, power supplies, cooling, and management into a single chassis. However, Arista's focus on open standards and programmable software gives it an edge in environments where customers demand flexibility and rapid innovation. The support for MRC is a prime example: it is an open protocol developed within the OCP, ensuring interoperability with a wide range of NICs and accelerators.

AI workloads are driving unprecedented demand for network bandwidth. Training large language models like GPT-4 requires thousands of GPUs or TPUs communicating over high-speed networks. The collective communication patterns—such as all-reduce and all-gather—create incast traffic that can overwhelm traditional Ethernet. To address this, Arista has incorporated advanced buffer management techniques in EOS, including dynamic threshold adjustment and per-flow queuing, to ensure that even during microbursts, packet loss is minimized.

Liquid cooling is becoming essential as power densities rise. The 7060XE7-64PRS-RV3-L is designed for direct liquid cooling using cold plates, which can dissipate heat from the ASIC and optics without relying on fans. This allows the switch to operate in environments where ambient temperatures are high or where fan noise must be minimized. Data center operators are increasingly adopting liquid cooling for AI racks, and Arista's offering fits neatly into that ecosystem.

In addition to hardware, Arista is investing heavily in AI network automation. EOS includes programmable APIs that allow orchestration tools like Kubernetes and OpenStack to automatically configure the network topology based on job requirements. The switches can be managed via Arista's CloudVision platform, which provides a single pane of glass for monitoring and provisioning across the entire data center fabric.

Source:Network World News