The CEO of Broadcom raises concerns about a significant change in the AI chip market

Broadcom, whose market value just crossed $1 trillion last week, has so far emerged as one of the biggest winners of the Q4 2024 earnings season.

As a pioneer in the design and production of semiconductor technology, Broadcom has had a fantastic year of consistent growth. With year-over-year gains of 118%, mostly due to advancements in artificial intelligence (AI) technology, AVGO stock is set to end 2024 on a very high note.

Even if Broadcom hasn't made as much money this year as Nvidia, the AI-chip industry could be about to undergo a big paradigm change that would greatly help the former. Hock Tan, the CEO of Broadcom, announced high revenue forecast for the next two years during last week's Q4 results call. This forecast reflects a developing trend in the AI-chip sector.

Nvidia has risen to the top of the rapidly expanding AI business over the last two years due to industry-wide demand. Large language models (LLMs) are the focus of many businesses, and Nvidia manufactures the only GPUs (graphic processing units) that are sufficiently powerful to support a wide range of AI systems.

In terms of processing power and efficiency, no firm has been able to create a GPU that can truly compete with Nvidia's up to this point. Tan did, however, address several factors during last week's Q4 results conference that point to a potential change in the AI-chip industry that would favor businesses producing specialized silicon chips, such as eXtreme Processing Units (XPUs), rather than Nvidia's general-purpose GPUs.

To put things in perspective, GPUs are said to be quite adaptable and are usually capable of handling a large range of processing jobs. The strong software ecosystem that surrounds Nvidia's CPUs has made them user-friendly.

XPUs, on the other hand, are extremely specialized and frequently designed for more specialized, high-computing jobs, usually related to artificial intelligence and machine learning. Although they might be more specialized, specialized chips appear to be outperforming more generic ones, according to Broadcom's Q4 earnings report.

As mentioned, Tan estimates that Broadcom's AI and AI networking will generate between $60 billion and $90 billion in 2027, a substantial increase from the $15 billion to $20 billion estimate made the previous year.