According to the Star Daily on December 16, last week's final trading day saw Broadcom soar 24% to a closing record high, making it the third semiconductor company in the world, after NVIDIA and Taiwan Semiconductor, with a market cap exceeding one trillion dollars.

The reason behind this is that Broadcom released its fourth-quarter Earnings Reports, showing semiconductor revenue of $8.23 billion, a year-on-year increase of 12%. Broadcom attributes this leap to the boom in AI infrastructure, stating that its AI revenue surged 220% this year to $12.2 billion, and will develop ASIC custom AI Chips with three very large clients in the future.

Melius Research Analyst Ben Reitzes believes this indicates that major companies will continue to invest in AI, which is beneficial for all AI Chip and network equipment manufacturers.

Currently, Broadcom's market cap stands at $1,049.9 billion, second among semiconductor companies only to NVIDIA at $3,287.8 billion. However, on the day of Broadcom's surge, NVIDIA's stock price fell. Mizuho Analyst Jordan Klein believes that Wall Street is focusing on the demand for ASICs from large tech companies, such as whether companies like Meta and Alphabet prefer using ASICs for generative AI applications over expensive GPUs, which could be one reason for NVIDIA's stock drop.

ASIC, short for Application-Specific Integrated Circuit, refers to chips specifically customized for certain needs, which in the current market context can refer to AI Chips. The Broadcom AI Chip mentioned earlier also falls under the category of ASICs. Compared to traditional general-purpose chips like CPUs and GPUs and semi-custom FPGA chips, ASICs have more targeted computational capabilities.

GTJA believes that currently, the AI ASIC single card computing power is lower than comparable GPU Chips, but due to their lower cost, they show a higher cost-performance ratio under commonly used accuracy for inference and have lower power consumption. Additionally, since ASICs are specifically designed for certain tasks, their computing power utilization rate might be higher; for example, Google's TPU utilization rate can exceed 50%. For cloud vendors, ASICs are also an important option to increase supply chain diversity.

Currently, the representative players in the ASIC market, Broadcom and Marvell (who together hold over 60% of the ASIC market share), have already outpaced the market significantly. The former, as the manufacturer of Google's self-developed AI chip TPU, has maintained a cooperative relationship for nearly ten years, and recent news shows that it is also collaborating with Apple to develop ASICs; the latter has designed over 2000 ASICs since launching this business 25 years ago, and has been invited by Amazon, Google, and Microsoft to develop custom AI chips.

In terms of growth potential, the future growth rate of ASIC is expected to exceed that of general-purpose accelerated computing chips. According to market forecasts, the scale of customized chips is expected to exceed 40 billion dollars by 2028, with a CAGR of 45%, while general-purpose accelerated computing chips are projected to reach a market size of 171.6 billion dollars, with a CAGR of 32%. The CEO of Broadcom boldly stated that by 2027, each customer will deploy 1 million AI Chips in their network clusters, and the demand for customized AI Chips in the market could reach between 60 billion to 90 billion dollars.

On December 15, Sinolink's Research Reports pointed out that there is a shortage of NVIDIA's Blackwell and GB200 Servers, and CSP is accelerating the mass production of ASICs, indicating that AI computing power remains strong. At the same time, Ethernet trends in AI applications are becoming increasingly evident. Continuing to be Bullish on the Industry Chain benefiting from AI computing power and AI Ethernet.

CSC believes that the demand and outlook for computing power remains strong, and the development of AI applications has exceeded expectations. Bullish on companies related to the Computing Power Sector, as large-scale customers are actively laying out ASICs, which will further open up market size, and attention should be paid to the industrial changes it may bring.