Broadcom launches new Tomahawk Ultra networking chip in AI battle against Nvidia

Broadcom has launched its Tomahawk Ultra networking processor, a new chip designed to accelerate AI data crunching by efficiently managing data flow between hundreds of chips within data centers. This product directly challenges Nvidia's NVLink Switch, offering a competitive edge by connecting four times more chips and utilizing an enhanced Ethernet protocol. Manufactured by TSMC and now shipping, the Tomahawk Ultra solidifies Broadcom's position as a key player in the burgeoning AI infrastructure market, further diversifying options beyond Nvidia's dominant offerings.

Analysis

Broadcom has launched its Tomahawk Ultra networking processor, a strategic move aimed directly at the high-growth artificial intelligence infrastructure market and a direct challenge to Nvidia's dominance. This new chip addresses a critical component of AI systems, managing data traffic between the hundreds of chips required for large-scale AI model training and inference. The Tomahawk Ultra's key competitive differentiator is its capacity to connect four times the number of chips as Nvidia's competing NVLink Switch, a significant advantage in 'scale-up' computing architectures. Furthermore, its reliance on an enhanced version of the open Ethernet standard, rather than a proprietary protocol, could appeal to data center operators seeking flexibility and avoiding vendor lock-in. The processor, now shipping, is manufactured by Taiwan Semiconductor Manufacturing on its 5-nanometer process, signaling a high-performance design. This launch builds upon Broadcom's existing credibility in the AI space, which includes its collaboration with Google on an alternative AI chip ecosystem, solidifying its position as a key enabler of AI hardware beyond Nvidia.

Key Decisions for Investors

• For Broadcom (AVGO) investors, the launch of the now-shipping Tomahawk Ultra provides a new, tangible catalyst in the high-margin AI networking segment; monitor for early customer adoption and revenue impact.
• Investors in Nvidia (NVDA) should consider this a material competitive threat to its data center networking business, as the Tomahawk Ultra's superior connectivity claims and open-standard approach could erode the appeal of its proprietary NVLink ecosystem.
• The announcement highlights that the AI hardware battleground is expanding beyond processors to interconnects, suggesting that a comprehensive portfolio assessment should include a company's networking fabric capabilities.
• The reliance on TSMC for 5nm manufacturing reinforces its pivotal role in the AI supply chain, benefiting from the intensified competition between hardware designers like Broadcom and Nvidia.