Enquire: Asia & Africa - +65-98008081 USA - +1-919-995-4114

Home Networking Broadcom Tomahawk 6 102.4Tbps 32-Port Platform

Broadcom Tomahawk 6 102.4Tbps 32-Port Platform

Brand:

The Broadcom Tomahawk 6 Platform (BCM78910) is a next-generation Ethernet switching platform built for AI, hyperscale cloud, and high-performance data-center fabrics. It delivers 102.4 Tb/s switching capacity in a single chip and supports 100G/200G SerDes, high-radix networking, advanced AI routing, and scale-up/scale-out architectures. The supplied configuration is identified as a 32-port Tomahawk 6 platform; exact port speed and physical implementation depend on the system vendor's design.

OVERVIEW

  • Broadcom Tomahawk 6 BCM78910 is a 102.4Tb/s Ethernet switching platform for AI clusters, hyperscale cloud and advanced data-center fabrics. Need: maximize bandwidth and reduce network tiers for large XPU deployments. Compatibility: system-vendor implementation determines ports, optics, breakout and software. Compare: Tomahawk 6 doubles Tomahawk 5's 51.2Tb/s switching capacity and supports 100G/200G SerDes. 

USE CASES

  • AI training clusters – Provides the switching bandwidth and AI-aware routing capabilities required for large distributed training fabrics.
  • AI inference infrastructure – Supports high-throughput Ethernet connectivity between accelerators, servers and network tiers.
  • GPU/XPU scale-out networks – High radix and 102.4Tb/s capacity can reduce the number of switching tiers required as accelerator clusters grow.
  • AI scale-up fabrics – Broadcom positions Tomahawk 6 for scale-up architectures supporting up to 512 XPUs in a supported configuration. 
  • Hyperscale cloud data centers – Provides high-capacity switching silicon for cloud operators building large Ethernet fabrics.
  • High-performance computing – Supports bandwidth-intensive distributed computing and parallel workloads.
  • Spine and leaf data-center networks – Provides high-density switching for modern Clos and other high-radix architectures.
  • Cloud-native infrastructure – Enables high-throughput Ethernet fabrics for compute, storage and east-west traffic.
  • Large-scale storage networks – Provides high-bandwidth switching for distributed storage and data-intensive applications.
  • Next-generation Ethernet AI fabrics – Supports emerging 100G/200G SerDes architectures and Ultra Ethernet-oriented deployments. 

KEY FEATURES

  • 102.4 Tb/s switching capacity – Provides extremely high aggregate bandwidth for large AI and cloud Ethernet fabrics. 
  • Broadcom BCM78910 architecture – Provides the core switching silicon for next-generation high-radix data-center platforms. 
  • 512 × 200G SerDes capability – Provides the high-speed electrical connectivity needed for dense 200G-per-lane architectures. 
  • 100G/200G SerDes support – Gives system designers flexibility for different AI and data-center interconnect implementations. 
  • High-radix switching – Allows system designers to connect more high-speed endpoints per switching tier, helping reduce fabric complexity.
  • Cognitive Routing 2.0 – Provides AI-oriented routing capabilities designed to improve network utilization and traffic distribution. 
  • AI scale-up support – Enables supported architectures targeting clusters of up to 512 XPUs.
  • AI scale-out support – Broadcom positions the platform for very large distributed AI networks, including 100,000+ XPUs in two-tier 200Gbps/link configurations. 
  • Ultra Ethernet Consortium compliance – Aligns the platform with modern Ethernet-based AI networking requirements. 
  • Long-reach passive copper support – Helps system designers extend high-speed connections while potentially reducing optical component requirements. 
  • Co-packaged optics option – Provides a path toward optical integration for high-bandwidth AI switching designs. 
  • Multiple Ethernet configurations – The BCM78910 family supports system implementations ranging from 512×200GbE to 64×1.6TbE, depending on platform design. 
  • Production availability – Broadcom announced production-volume shipping of the Tomahawk 6 family in March 2026, making it relevant for current next-generation AI infrastructure planning.

TECHNICAL SPECIFICATIONS

  • Brand: Broadcom 
  • Product family: Broadcom Tomahawk 6
  • Platform / SKU: Tomahawk-6-Platform
  • Broadcom ASIC: BCM78910
  • Product type: Ethernet switch ASIC / platform
  • Switching capacity: 102.4 Tb/s
  • Architecture: High-radix Ethernet switching
  • SerDes: 100G / 200G
  • SerDes lanes: 512 × 200G
  • Official example configuration: 64 × 1.6TbE
  • Official example configuration: 128 × 800GbE
  • Official example configuration: 256 × 400GbE
  • Official example configuration: 512 × 200GbE
  • Supplied platform configuration: 32 ports
  • Exact port speed: System-vendor dependent
  • AI scale-up: Up to 512 XPUs in supported architecture
  • AI scale-out: 100,000+ XPUs at 200G/link in supported two-tier architecture
  • Routing Cognitive Routing: 2.0
  • AI networking: AI-optimized routing and congestion-management features
  • Ethernet ecosystem: Ultra Ethernet Consortium compliant
  • Copper support: Long-reach passive copper
  • Deployment: AI, HPC, hyperscale cloud, data center
  • Compatibility: System-vendor platform dependent
  • Predecessor: Tomahawk 5 / 51.2Tb/s
  • Bandwidth vs. Tomahawk 5: 2× aggregate switching capacity

Why Choose This Product?

  • 102.4Tb/s switching capacity for extreme AI and cloud networking workloads
  • 2× the bandwidth of Tomahawk 5 for a major generational capacity increase
  • 200G SerDes for next-generation high-speed Ethernet designs
  • High-radix architecture helps reduce network tiers in large AI fabrics
  • AI-optimized routing improves traffic management for demanding accelerator workloads
  • Scale-up and scale-out support provides architecture flexibility
  • CPO option creates a path toward high-bandwidth optical integration
  • Production-volume availability makes Tomahawk 6 relevant to current AI infrastructure deployments

CUSTOMER REVIEW

No Reviews Found

FAQ

No FAQs Found