Arista Networks has launched the 7060XE7 Series, a notable portfolio of 1.6T networking platforms tailored for AI infrastructure. This series marks a shift toward rack-scale systems designed to meet the high demands of artificial intelligence workloads, highlighting the need for greater bandwidth and power efficiency.
The new 7060XE7 family features both fixed switch platforms and configurable rack-scale systems, accommodating various vertical and horizontal AI workflows. All systems will run on Arista’s Extensible Operating System (EOS), which is crafted to handle the challenging microbursts typical of AI communication. The incorporation of intelligent packet buffering within EOS is particularly significant, as it effectively manages the dynamic nature of AI data traffic.
Built on Broadcom’s advanced Tomahawk 6 silicon, the 7060XE7 Series aims to enable scale-out AI fabrics. Additionally, Arista is working with AMD on next-generation compute silicon and network interface cards (NICs), ensuring that their products remain at the forefront of technology. This strategic partnership underscores Arista’s dedication to delivering solutions capable of efficiently managing the extreme density, power, and thermal management required by modern AI applications.
The 7060XE7 Series provides various configurations to meet diverse operational needs. Key models include the 7060XE7-64PS and 7060XE7-64PRS, both expected to launch in Q4. These air-cooled systems support pluggable Integrated Heat Sink (IHS) and Riding Heat Sink (RHS) optics, allowing for adaptability in data center environments. The RHS configuration specifically targets future liquid-cooled AI infrastructures, reflecting the evolving requirements for high port density and thermal efficiency.
For high-density clusters, the 7060XE7-64PRS-RV3-L, a specialized 2OU liquid-cooled platform, is set to debut in Q1 2027. This model boasts 224G SerDes and integrates seamlessly with liquid-cooled XPU servers, maximizing power efficiency by eliminating internal fans. Another significant addition is the 7060XE7-128PE, which will feature 128 800G ports in an air-cooled 4RU design, also scheduled for release in Q1 2027. This model is tailored for environments that require flexible deployment and compatibility with existing infrastructure.
On the software side, EOS stands out as the primary network operating system while also supporting open-source platforms like Sonic and OpenSwitch. This flexibility allows users to customize their systems according to specific needs while ensuring compatibility with various operational environments.
Arista’s new offerings represent a crucial advancement in tackling the challenges of scaling AI infrastructures. The company’s emphasis on higher bandwidth, enhanced power efficiency, and tighter integration across compute, optics, silicon, cooling, and network software signifies a clear direction for the future of AI technologies. As the demand for AI capabilities continues to grow, Arista's innovative switch family is well-positioned to support the next generation of AI-driven applications.
The stories that move AI & crypto markets — before the market reacts.
Free. 7am ET. Five stories. 62,400 readers.
