Arista Networks has unveiled its 7060XE7 Series, a new family of 1.6T networking platforms designed to provide the foundation for rack-scale AI infrastructure. The portfolio represents a strategic shift from standalone high-performance switches to integrated rack-scale systems that address the extreme demands of modern AI workloads, including high bandwidth, low latency, and efficient power and thermal management.
The 7060XE7 family includes fixed switch platforms and configurable rack-scale systems, all powered by Arista's Extensible Operating System (EOS). EOS brings advanced features such as low-latency forwarding and intelligent packet buffering to handle the intense microbursts common in AI communication and collective patterns. The switches are built on Broadcom's Tomahawk 6 silicon, which delivers 1.6T of switching capacity per chip, enabling higher port densities and faster data movement.
AI-Optimized Hardware Configurations
Arista has announced several specific models within the 7060XE7 Series, catering to different deployment scenarios. The 7060XE7-64PS and 7060XE7-64PRS are 4U air-cooled rack switches available in Q4 2026. They support 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 and extreme port density requirements.
The 7060XE7-64PRS-RV3-L is a specialized 2OU liquid-cooled platform for high-density clusters. It features 224G SerDes and uses DC power from the ORv3 rack standard. This system contains no internal fans, integrating directly with liquid-cooled XPU servers to maximize power efficiency. It will be available in Q1 2027.
For environments requiring deployment flexibility and backward compatibility, Arista offers the 7060XE7-128PE, which provides 128 800G ports in an air-cooled 4RU design using 100G SerDes. This model is also slated for Q1 2027 availability. These configurations underscore Arista's commitment to supporting both current cooling methods and next-generation liquid-cooled architectures as AI clusters grow denser.
Software and Open Standards
EOS remains the featured network operating system across the 7060XE7 family, but Arista also supports open-source alternatives such as SONiC (Software for Open Networking in the Cloud) and OpenSwitch. This flexibility allows customers to choose the software stack that best fits their operational models. A key software innovation 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 use multiple network paths over Ethernet, greatly improving fabric utilization and load balancing.
As described by Arista's CTO Kenneth Duda and distinguished engineer Alan Judge, MRC allows endstation NICs to stripe traffic across multiple links and paths, automatically handling out-of-order packets. The protocol responds to congestion signals such as ECN and packet trimming, shifting load to the best-performing paths and avoiding failed links. Arista has proven in production that MRC achieves very high fabric utilization while interoperating seamlessly with scale-across and WAN networks using standard dynamic routing protocols.
The software also includes advanced load balancing, congestion management, telemetry, and diagnostics tools that are essential for AI networking. These capabilities help network operators monitor and optimize performance in real time, ensuring that AI workloads receive the deterministic, low-latency connectivity they require.
Ecosystem and Competitive Landscape
Arista's new switch family enters a competitive ecosystem where vendors like Cisco, Nvidia, and Celestica are also developing 1.6T Ethernet solutions. The 7060XE7 Series benefits from strong customer and ecosystem validation, with endorsements from major cloud providers such as Microsoft Azure, Oracle Cloud Infrastructure, and Meta, as well as silicon partners AMD and Broadcom. Arista is also collaborating with AMD on next-generation compute silicon and NICs to enable scale-out AI fabrics, further integrating the networking layer with the compute layer.
The shift to rack-scale systems reflects a broader industry trend. According to Sameh Boujelbene, vice president of data center switch and AI networks market research at Dell'Oro Group, the new Arista portfolio signals where large-scale AI fabrics are headed: higher bandwidth, better power efficiency, and tighter integration between compute, optics, silicon, cooling, and network operating software. This holistic approach reduces total cost of ownership and accelerates time to deployment for AI clusters.
Technical Deep Dive: Tomahawk 6 and Beyond
Broadcom's Tomahawk 6 silicon is a key enabler of the 7060XE7 Series. This switch chip delivers 1.6T of bandwidth per device, supporting 64 ports of 800G or 128 ports of 400G in various configurations. The high serial rate and low power per bit make it well-suited for AI back-end networks, where massive parallel traffic patterns demand both throughput and efficiency. Arista's implementation includes advanced buffering and scheduling capabilities to handle microbursts and incast patterns that can occur during distributed training of large language models.
Beyond the initial platforms, Arista is likely to extend the 7060XE7 family with additional form factors and software enhancements. The integration of MRC and support for open protocols positions these switches as foundational components in AI fabrics that must scale from a single rack to thousands of nodes. The liquid-cooled model in particular points to a future where entire data center pods operate without traditional air handling, reducing energy overhead and enabling higher compute densities.
Arista's roadmap also includes continued work with AMD on next-generation NICs and processors. These partnerships aim to optimize the entire data path from GPU to switch to storage, ensuring that AI training and inference workloads encounter minimal bottlenecks. The combination of high-performance switching, intelligent software, and open collaboration makes the 7060XE7 Series a compelling choice for organizations building dedicated AI infrastructure.
Implications for AI Data Centers
The 7060XE7 Series arrives at a time when AI workloads are reshaping data center network design. Traditional leaf-spine architectures are giving way to more specialized topologies that support both scale-up (within a rack) and scale-out (across racks) communication patterns. Arista's new switches are explicitly designed for these environments, with high port density to connect many GPU servers and low latency to minimize training times.
Support for MRC is particularly significant because it addresses one of the most persistent challenges in AI networks: load imbalance. By using multiple paths simultaneously and dynamically steering traffic based on congestion, MRC can improve effective fabric utilization from around 50% to over 90%, according to Arista's production data. This translates to faster completion times for distributed training jobs and higher return on investment for expensive GPU clusters.
Additionally, the thermal management features of the liquid-cooled model open the door to more efficient cooling strategies. As AI accelerators approach 1000W per chip, liquid cooling becomes essential to remove heat without wasting energy on fans. Arista's integration with ORv3 power standards ensures compatibility with open rack architectures, making it easier for customers to adopt standardized infrastructure.
The competitive landscape continues to heat up, with Cisco and Nvidia also pushing 1.6T offerings. However, Arista's strong relationships with hyperscalers and its open ecosystem approach give it an edge in environments that require interoperability with existing SONiC-based deployments or with multiple vendor NICs. The company's longevity in the data center switching market, built on a foundation of software reliability and customer support, adds to its credibility.
As AI models grow larger and training clusters expand to tens of thousands of accelerators, the networking backbone must keep pace. Arista's 7060XE7 Series represents a timely and robust response, combining cutting-edge silicon, intelligent software, and flexible form factors to meet the needs of next-generation AI infrastructure. The availability of air-cooled models in late 2026 and liquid-cooled options in early 2027 gives customers a clear migration path toward more efficient, higher-performing networks.
Source: Network World News