Broadcom is introducing a new networking chip designed to assist companies in constructing artificial intelligence computing systems by connecting numerous chips that process data, intensifying its competition with Nvidia.

The Thor Ultra chip allows computing infrastructure managers to implement significantly more chips than they could normally, enabling them to create and operate the extensive models that drive AI applications like ChatGPT.

The Thor Ultra will compete with Nvidia's networking interface chips and seek to solidify Broadcom's dominance in network communications within data centers tailored for AI uses.

This follows Broadcom’s announcement of an agreement to deliver 10 gigawatts of specialized chips for OpenAI, the creator of ChatGPT, starting in the latter half of 2026, putting pressure on Nvidia's dominance in the AI accelerator industry.

AI signifies a significant opportunity for Broadcom. CEO Hock Tan mentioned late last year that the market targeted by the company for its diverse AI chips is projected to be between $60 billion and $90 billion by 2027, with segments for its networking chips and the data center processors it aids Alphabet's Google and OpenAI in producing.

Also Read: AI Appreciation Day: From Innovation to Transformation

Broadcom announced AI revenue of $12.2 billion for fiscal year 2024. In September, it revealed a new, unnamed client worth $10 billion for its specialized data center AI chips.

Also Read: Karnataka to Become Quantum Capital of Asia Soon

The Thor Ultra chip, a component of Broadcom's expanding collection of networking chips, serves as an essential connection between an AI system and the remainder of the data center. The networking chips assist data center managers in transferring data within a facility.

Also Read: The Global Fintech Fest 2025: Enabling Finance for Better World

Ram Velaga, a senior vice president at Broadcom, states, "In a distributed computing system, the network is crucial for constructing these sizable clusters. Therefore, I'm not shocked that anyone involved in the GPU industry aims to ensure their involvement in networking.