MENU

Celestica DS6005: 102.4Tbps CPO Switch Specs for AI Data Centers

Unite.AI USA
Overview
Celestica has introduced the DS6005, a fully liquid-cooled co-packaged optics (CPO) switch delivering 102.4Tbps capacity for AI and machine learning networks. Built on Broadcom’s Tomahawk 6 – Davisson silicon, the system reduces power consumption and latency compared to traditional pluggable optics. This platform addresses the critical bandwidth and thermal constraints of next-generation AI scale-up and scale-out fabrics.
In Depth

Key Findings

The Celestica DS6005 is a high-capacity switching system designed for AI data centers, providing 102.4Tbps of switching capacity. The system utilizes co-packaged optics (CPO) to place optical engines directly on the switch ASIC package, eliminating conventional front-panel pluggable optics to optimize bandwidth density and power efficiency. The DS6005 is fully liquid-cooled and housed in a 1.5OU enclosure compliant with OCP ORv3-HPR rack specifications.

Technical Details

The hardware is built around the Broadcom Tomahawk 6 – Davisson switch chip and an AMD Ryzen x86 CPU. Key technical specifications include:

  • Capacity: 102.4Tbps total switching capacity via 64 ports of 1.6TbE.
  • Connectivity: 64 optical interconnects utilizing 16x 4x MMC-16 connectors.
  • Interfaces: 16 ELSFP interfaces for field-replaceable external laser modules, one 100GbE QSFP28 port, RJ45 console, out-of-band GbE management, one USB 3.0 Type A port, and one USB-C console port.
  • Software: Support for SONiC to enable an open architecture.

Background and Context

The DS6005 leverages Broadcom’s Tomahawk 6 – Davisson silicon, which Broadcom began shipping on October 8, 2025. This third-generation CPO Ethernet switch integrates 16 Davisson DR optical engines (6.4Tbps each) using TSMC’s COUPE technology and advanced substrate-level multi-chip packaging. Broadcom reports that this integration reduces optical interconnect power consumption by 70%, which is more than 3.5 times lower than traditional pluggable solutions. The chip supports scale-up clusters of 512 XPUs and two-tier networks of over 100,000 XPUs at 200Gbps per link.

Significance and Outlook

The system is optimized for high-performance computing, frontier cloud and enterprise AI models, and hyperscale infrastructure. Industry analysts from Dell’Oro Group and 650 Group note that AI networking constraints are shifting from compute to physical infrastructure, specifically power and density. The transition of CPO from an industry theme to production-ready systems like the DS6005 provides a path to address these constraints in high-density scale-out fabrics.

Source: https://www.unite.ai/celestica-unveils-ds6005-co-packaged-optics-switch-for-ai-data-centers/

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.

  • Your email and selected fields are used only to deliver the newsletter.
  • We never share your information with third parties.
  • You can unsubscribe anytime via the link in each email.

See our Privacy Policy for details.

Takes about a minute · Unsubscribe anytime

Published by Troy-Technical, an independent site run by one engineer with a career in materials development.
About the author / Contact info@troy-technical.jp
Let's share this post !

Author of this article

TOC