Arista Unveils 1.6T Rack-Scale Switch Family for AI Infrastructure

Why It Matters

The 7060XE7 gives AI workloads the bandwidth, power efficiency, and congestion‑aware networking needed for next‑gen clusters, positioning Arista as a key supplier in a rapidly expanding market.

Key Takeaways

• Arista's 7060XE7 offers up to 1.6 Tbps per rack
• Supports both air‑cooled and fan‑less liquid‑cooled designs
• MRC protocol enables dynamic multipath load balancing for AI traffic
• Broadcom Tomahawk 6 silicon powers 800G and 128‑port models
• Early adopters include Azure, Oracle Cloud, Meta, and AMD

Pulse Analysis

The AI boom is driving data centers toward 1.6‑terabit Ethernet, a bandwidth tier that can sustain the massive tensor‑core traffic of modern models. Competitors such as Cisco, Nvidia and Celestica have already introduced high‑speed fabrics, but the market remains fragmented, with many customers still piecing together disparate switches, optics and cooling solutions. Arista’s entry with the 7060XE7 family consolidates these components into a single rack‑scale offering, promising tighter integration and simplified procurement for hyperscalers and enterprise clouds.

At the hardware level, the 7060XE7 leverages Broadcom’s Tomahawk 6 ASIC, delivering up to 1.6 Tbps of line rate across 64‑port 800 GbE or 128‑port configurations. The lineup spans traditional air‑cooled 4U switches, a liquid‑cooled 2‑OU module that eliminates internal fans, and a hybrid design that can accommodate both Integrated Heat Sink (IHS) and Riding Heat Sink (RHS) optics. Running Arista’s Extensible Operating System (EOS), the switches also support open‑source stacks like SONIC and OpenSwitch, while the inclusion of the Open Compute Project’s Multipath Reliable Connection (MRC) protocol provides real‑time congestion awareness and automatic path steering—critical for AI workloads that generate microbursts and demand near‑zero latency.

From a business perspective, the 7060XE7’s early validation by Microsoft Azure, Oracle Cloud, Meta, AMD and Broadcom signals strong market traction. By moving from isolated high‑performance switches to fully integrated rack‑scale systems, Arista can capture a larger share of the AI‑fabric spend, which analysts estimate will exceed $30 billion by 2028. The combination of power‑efficient cooling, open‑software flexibility, and proven congestion‑management positions Arista to become a preferred networking partner for the next wave of AI‑driven cloud services.