AI concept stocks performed strongly in Wednesday’s market trading, with Broadcom (AVGO) hitting a record high. The company’s stock rose 3.5% on the day, significantly outperforming the S&P 500 (flat) and the Nasdaq Composite (up 0.5%). By the end of 2025, Broadcom’s stock had accumulated a gain of 66.5%. The increase in its valuation is largely attributed to positive news surrounding the collaboration between NVIDIA and Oracle to build a new supercomputer.
NVIDIA announced it will collaborate with Oracle to build a supercomputer equipped with 100,000 Blackwell GPUs for the U.S. Department of Energy, dedicated to scientific research tasks. Although the announcement did not directly mention Broadcom, this major project is seen as a positive catalyst for Broadcom’s demand growth. As an expert in connectivity chips and software solutions, Broadcom produces AI data center products that are closely linked to NVIDIA’s ecosystem. Historical evidence suggests that when NVIDIA secures major contracts or gains growth momentum, Broadcom typically receives related orders in the future, creating a synergistic effect.
Broadcom holds a leading position in the fields of custom chips (ASIC) and networking components. Although its current AI revenue is significantly lower than NVIDIA’s ($5.2 billion vs. $41.0 billion in the most recent fiscal quarter, respectively), it exhibits strong growth momentum. Broadcom’s AI revenue grew 63% year-over-year, surpassing the 56% growth of NVIDIA’s data center business. As the demand for AI workloads shifts from model training to inference applications, Broadcom’s custom processors demonstrate unique advantages. AMD CEO Dr. Lisa Su pointed out that the growth rate of AI inference demand has already surpassed that of training demand, especially in verticals like automotive, healthcare, and finance, where the explosion in the number of specialized models is driving demand for Broadcom’s custom processors.
Compared to NVIDIA’s general-purpose GPUs, Broadcom’s custom processors are designed for specific tasks, offering lower power consumption and higher performance. This technological advantage has made them popular with many leading customers. For example, OpenAI, which relied on NVIDIA to train ChatGPT, has partnered with Broadcom and plans to deploy 10 gigawatts of custom AI processors between 2026 and 2029. Analysts estimate this collaboration could generate approximately $100 billion in revenue for Broadcom. Furthermore, Broadcom’s revenue backlog at the end of the last fiscal quarter was as high as $110 billion, far exceeding its actual revenue of $60 billion over the past year, reflecting a strong increase in order backlogs.
Broadcom’s estimated market share in the custom AI processor segment is around 70%. Management expects its share of the overall AI chip market to increase from 11% in 2025 to 24% in 2027. Although the current P/E ratio is as high as 87, its forward P/E ratio is 37, indicating market confidence in its profit growth. According to data, Broadcom’s PEG ratio is only 0.55, below the threshold of 1, suggesting that based on the expected annualized growth over the next five years, its stock is still undervalued. Investors should watch Broadcom’s expansion potential in the AI inference chip market – it is estimated that by 2030, the share of ASICs in chips performing AI inference tasks will rise from 15% to 80%, and Broadcom is a core beneficiary of this trend.
