TLDR
- Broadcom unveiled Thor Ultra networking chip Tuesday to compete with Nvidia in AI infrastructure market
- OpenAI deal worth $10 billion for custom chips begins second half of 2026
- Company posted $12.2 billion AI revenue in fiscal 2024, targets $60-90 billion market by 2027
- Thor Ultra doubles bandwidth of previous version to support massive AI computing clusters
- Bernstein keeps Outperform rating with $400 price target on strong AI growth outlook
Broadcom introduced its Thor Ultra networking chip Tuesday. The processor enables data centers to build larger AI computing systems by connecting hundreds of thousands of chips.
The launch comes one day after Broadcom announced a major OpenAI partnership. The company will deliver 10 gigawatts of custom chips starting in the second half of 2026.
Thor Ultra directly challenges Nvidia’s networking interface technology. The chip helps data center operators move information around facilities running large language models like ChatGPT.
AI Market Expansion Plans
CEO Hock Tan previously outlined the company’s AI chip ambitions. Broadcom is pursuing a $60 billion to $90 billion market opportunity by 2027.
The company generated $12.2 billion in AI revenue during fiscal 2024. September brought news of another unnamed $10 billion customer for custom data center chips.
Broadcom’s networking chips serve as critical links in AI infrastructure. They connect computing clusters to broader data center systems.
“In the distributed computing system, network plays an extremely important role in building these large clusters,” said Ram Velaga, Broadcom senior vice president.
Engineers doubled Thor Ultra’s bandwidth compared to the prior version. The development team conducted extensive testing throughout early production stages.
Partnership Strategy and Revenue Growth
Broadcom focuses on chip design rather than selling complete servers. The company provides reference designs that customers use to build networking infrastructure.
“For every dollar we invest in our silicon, there is at least $6 to $10 that our ecosystem partners are investing,” Velaga explained.
The approach has proven lucrative with clients like Google. Broadcom has worked on multiple generations of Google’s Tensor processor over the past decade.
Bernstein analyst Stacy Rasgon maintained an Outperform rating Tuesday. The firm set a $400 price target on expectations of positive earnings revisions.
Rasgon believes current revenue estimates are too conservative. He addressed margin concerns by stating he doesn’t expect gross profits to “plummet” despite AI expansion.
Broadcom maintains 77.19 percent gross profit margins. The semiconductor giant has a $1.68 trillion market cap with 28 percent year-over-year revenue growth.
The stock gained 9.88 percent to close at $356.70 on October 13. Pre-market trading on October 14 showed shares at $347.30.
Custom Chip Business Grows
Broadcom helps design custom AI chips for major cloud computing companies. These processors have generated billions in revenue based on analyst estimates.
The company’s San Jose labs house engineers working on current and future networking chip designs. Teams collaborate on packaging decisions, power requirements, and thermal management.
The Thor Ultra chip joins Broadcom’s expanding catalog of networking processors. The company also produces the Tomahawk series of networking switches for data centers.
Broadcom’s OpenAI partnership represents another win in the custom chip market. The deal follows the company’s track record with Google’s Tensor processors and other unnamed hyperscale customers.
