Lisa Su, chief executive officer (CEO) of AMD, is facing complex circumstances. Amid the ongoing dominance of competitor NVIDIA in the general-purpose artificial intelligence (AI) semiconductor market, the growth of Broadcom in the customized semiconductor (ASIC) market is significant as it emerges as an alternative to AI semiconductors. There are also concerns that the demand for high-performance general-purpose AI semiconductors may decline due to the impact of the Chinese AI company DeepSeek, which could affect AMD's future performance.
On the 5th, foreign media outlets including Reuters pointed out that AMD's growth could slow due to NVIDIA's market dominance and the development of proprietary customized semiconductors by big tech corporations such as Microsoft (MS), Amazon, and Meta. Ryuta Makino, an analyst at Gabelli Funds, noted, "Investors expect the future AI semiconductor market to be led by NVIDIA and the customized semiconductor market."
AMD has also shown remarkable performance in the AI server semiconductor market, in addition to its traditional focus on central processing units (CPUs), rapidly rising as a contender against NVIDIA. AMD's revenue in the data center chips sector in the fourth quarter of last year was approximately $3.86 billion (about 5.6 trillion won), up about 69% year-on-year, accounting for more than half of total revenue.
However, despite this growth, NVIDIA's monopoly in the AI semiconductor market remains unbroken. Reuters stated, "NVIDIA's monopoly structure is still solid, and its CUDA software is widely used as a standard by developers," adding that "the high costs associated with changing chip suppliers continue to pose significant barriers for AMD entering the market."
The fact that global big tech corporations are entering the market to create customized semiconductors optimized for their data centers is also unwelcome news for AMD. MS, Amazon, and Meta revealed customized semiconductors optimized for their data center operations last year. An industry insider stated, "Big tech companies operating data centers are AMD's main clients, but there is strong demand to break away from their monopoly structure," adding that "using general-purpose AI semiconductors not only incurs tremendous expenses but also carries the risk of inefficiencies in power consumption and performance, prompting companies to seek semiconductors optimized for their key application areas."
In response, Broadcom and Marvel Technology, which provide customized semiconductors in collaboration with big tech companies, are experiencing soaring business. Hock Tan, CEO of Broadcom, mentioned that the company expects to achieve revenues of $90 billion (about 131 trillion won) in the AI sector by 2027. Reflecting such market expectations, Broadcom's stock price more than doubled last year, while Marvel's increased by approximately 83%. In contrast, AMD's stock price had fallen about 18% as of the end of last year compared to the beginning of the year.
The impact of DeepSeek's emergence cannot be ignored. DeepSeek has been evaluated as demonstrating superior performance at lower costs compared to existing AI services such as ChatGPT, raising concerns that demand for high-performance AI semiconductors could shrink. Chan Min-sook, a researcher at Korea Investment & Securities, explained, "The fact that the DeepSeek model achieved the same performance as big tech companies at one-tenth of the cost is leading to concerns that it could reduce demand for semiconductors, including NVIDIA's graphics processing units (GPUs)."
An industry insider stated, "It is a fact that Lisa Su took the helm of AMD, which was facing over $1 billion in annual operating losses in 2014, and revived the company." However, they also noted that "faced with the situation where competitors like NVIDIA and Broadcom are taking the lead in the AI business, which has emerged as a crucial battleground for semiconductor corporations, AMD needs to find new breakthroughs."