Broadcom has officially begun shipping its latest networking chip, the Tomahawk 6, designed to supercharge artificial intelligence (AI) infrastructure by delivering double the performance of its predecessor.
Broadcom Tomahawk 6 Delivers Breakthrough in AI Networking Performance
Unveiled on Tuesday, the Tomahawk 6 networking chip marks a major leap forward in AI data centre networking technology. According to Ram Velaga, Senior Vice President at Broadcom, the new chip offers twice the performance of previous generations, enabling more efficient and scalable deployment of AI-driven workloads.
Velaga told Reuters that the Tomahawk 6’s speed enhancements allow data centre architects to use fewer switches to achieve the same computing performance, significantly reducing infrastructure complexity and cost.
Designed for Massive-Scale AI Clusters
With AI workloads demanding ever-larger data clusters, chips like Tomahawk 6 are critical for scaling. Velaga noted that the chip is built to support AI clusters exceeding 100,000 GPUs and anticipates future facilities housing up to 1 million GPUs within a single physical building.
Ethernet-Based Technology Competes with InfiniBand
Broadcom’s Tomahawk series relies on Ethernet, the long-standing industry standard, rather than more specialised networking technologies like Nvidia’s InfiniBand.
“All of these networks can be very simply done on Ethernet. You don’t need esoteric technologies,” Velaga emphasized.
This highlights Broadcom’s commitment to standardisation, which allows data centres to scale faster and more affordably compared to proprietary alternatives.
First Tomahawk Chip to Use Chiplets
The Tomahawk 6 is also Broadcom’s first networking chip to use chiplet-based architecture, a design technique popularised by companies like AMD. By combining several smaller chips into a single package, the Tomahawk 6 doubles the silicon area, improving efficiency and modularity.
Built on TSMC’s 3nm Process
The chip is manufactured using Taiwan Semiconductor Manufacturing Company’s (TSMC) advanced 3nm process, ensuring state-of-the-art power efficiency and performance. This move positions Broadcom as a key player in next-generation semiconductor development for AI infrastructure.
Key Features of Broadcom Tomahawk 6
- Double the performance of previous Tomahawk chips
- Optimized for AI data centers with 100,000+ GPUs
- Based on Ethernet, not InfiniBand
- Chiplet architecture for scalability
- Built on TSMC’s 3nm process technology
Broadcom Poised to Lead in AI Networking Infrastructure
As global demand for AI infrastructure grows, Broadcom’s Tomahawk 6 networking chip is set to become a cornerstone in powering next-generation AI data centres. With cutting-edge performance, scalable chiplet design, and Ethernet compatibility, it offers a compelling alternative to proprietary technologies in the race to build massive AI clusters.
