Report: Broadcom will launch new networking chips to challenge NVIDIA's dominance

According to media reports on Monday, Broadcom released the Thor Ultra network chip, aimed at helping enterprises build large-scale AI computing systems, further intensifying competition with NVIDIA in the AI infrastructure sector. The chip can support the connection of hundreds of thousands of chips, providing underlying network support for AI applications such as ChatGPT.

Broadcom CEO Hock Tan previously stated that the company expects the AI chip market opportunity to reach between $60 billion and $90 billion by 2027, covering both network chips and data center processors. This latest product release comes as Broadcom accelerates its layout in the AI market.

The chip release follows a significant cooperation agreement between Broadcom and OpenAI. It is reported that Broadcom will provide OpenAI with customized chips worth $10 billion starting in the second half of 2026, directly challenging NVIDIA's dominance in the AI accelerator market.

The launch of the new chip will strengthen Broadcom's control in the AI data center network communication field and create direct competition with NVIDIA's network interface chips.

AI Market Prospects Are Bright, Broadcom's Revenue Grows Rapidly

Broadcom has already shown strong growth momentum in the AI field, with AI-related revenue reaching $12.2 billion in fiscal year 2024. In September, it announced a new customer order worth $10 billion for customized data center AI chips.

Broadcom Senior Vice President Ram Velaga stated in an interview:

In distributed computing systems, networks play an extremely important role in building large clusters. It is not surprising that any company engaged in GPU business wants to ensure its participation in network business.

The company expects the target market size in 2027 to be distributed between network chips and data center processors, with an overall size reaching $60 billion to $90 billion.

Technological Upgrades Support Large-Scale AI Deployment

The Thor Ultra chip, as an important part of Broadcom's network chip product line, plays a key connecting role between AI systems and other parts of data centers. This chip enables computing infrastructure operators to deploy a number of chips far exceeding previous scales, thereby building and operating large models that support AI applications.

Broadcom engineers achieved a twofold bandwidth increase on the Thor Ultra compared to the previous version. From the early stages of production, the chip has undergone rigorous testing and evaluation processes.

When manufacturing the Thor Ultra or flagship products like the Tomahawk network switch series, engineers need to build a complete system around the chip. Velaga stated that engineers discuss key parameters such as chip packaging types, power consumption requirements, and heat dissipation with the hardware systems team.

Design-Oriented Strategy, Collaborating with Ecosystem Partners

Although Broadcom itself does not sell servers, it provides customers with components and system design solutions to reference for building their network infrastructure.

Velaga revealed the company's investment strategy:

For every dollar we invest in chips, ecosystem partners invest at least $6 to $10. Therefore, our focus is on design, just as we are almost ready to go into production. **

Broadcom's collaboration with major cloud computing companies has begun to show results. The company has participated in multiple generations of design work for Google's Tensor processors, a project that started over a decade ago. According to analysts' estimates, Tensor chips have generated billions of dollars in revenue for Broadcom