Master industry analysis to capture investment trends.
With the explosion of AI, how should the chip industry invest and lay out?
From January 1st to September 14th, 2023, the NASDAQ index in the U.S. rose nearly 30%. This increase was mainly driven by seven giant companies: Apple, Microsoft, Google, Amazon, Tesla, Meta, and Nvidia, all of which had better performance than the NASDAQ, with Nvidia having over twice the highest increase.
The common feature of these companies is that each of their businesses is more or less related to artificial intelligence (AI). For example, Apple's Vision Pro requires AI technology, Microsoft's OpenAI is the engine of this wave of AI, and Nvidia is a supplier of AI computing chips.
At the beginning of 2023, after the popularity of ChatGPT, Nvidia CEO Huang Renxun proclaimed boldly that the AI era of iPhone has arrived.
More than a decade ago, the birth of the Apple phone was a key moment in the development history of the mobile internet industry. Is the emergence of ChatGPT really the moment of AI's iPhone? There may not be a definite answer to this question at the moment, but judging from the stock trends of AI-related companies, it is an indication that the market may temporarily believe it.
The development of AI will benefit many industries. For example, we can see that in the Futubull industry chain function, Chatgpt covers a wide range of upstream and downstream industries, and many listed companies participate in it.
Behind these industries, there is one industry that cannot be avoided, and that is the chip industry, especially the computing power chip. Because whether it is the development of large AI models or their subsequent applications, they rely on the support of computing power.
In this article, we will focus on the chip industry analysis, including:
1. The overall development of the chip industry, 2. The classification and industrial chain of the chip industry, 3. Some characteristics of the chip industry, and 4. The investment logic of the chip industry.
1. The overall development of the chip industry
Warren Buffett once said, "Life is like a snowball, the important thing is finding very wet snow and a very long hill." Applied to investment, it means finding a long-term and reasonably growing track. What is the market space and growth rate of the chip industry?
Today, humans have entered the digital economy era, and chips are widely used in all aspects of life. From small things such as mobile phones, headphones, and smart watches, to large things such as computers, televisions, and even automobiles, they all use chips. If all technology-related industries are compared to a gold mine, then chips are the shovel for digging gold. To dig gold, you must use this shovel.
According to institutional reports, the global semiconductor chip industry reached a market size of over $574 billion in 2022, more than four times its size 21 years ago, with a compound annual growth rate of approximately 7% during that time.
Institutions predict that the chip industry will still maintain an annual compound growth rate of about 7%, and its market size will exceed $1 trillion in 2030, becoming a super large track with huge space.
2. The classification and industrial chain of the chip industry
For such a large industry as chips, in order to invest effectively, we also need to find the most potentially profitable sub-industry. At the same time, we also need to understand the division of labor in its industrial chain in order to better understand its investment logic.
From a classification perspective, all chips can be divided into eight types. For example, logic chips mainly include CPUs and GPUs that we often hear about, which are computer processors and graphics cards respectively; storage chips, including memory frequently used in computers and mobile phones; and analog chips, mainly used in the communication field, such as 5G analog chips in mobile phones. Among them, the largest type is logic chips, accounting for more than 30% of the entire chip industry.
CPUs in logic chips are adept at complex serial computing, while GPUs, originally known as graphics cards, are good at simple parallel computing, which is exactly what AI development requires and what GPUs can provide: simple, massive computing power. Therefore, GPUs have become the main computing chip, and institutions predict that their future growth rate may exceed 30%, making it a subdivision field worth focusing on due to its fast growth and large space. Product structure, 10-30 billion yuan products operating income of 401/1288/60 million yuan respectively.
Looking at the industry chain of the chip industry, a chip generally goes through 3 stages from scratch: design, manufacturing, and encapsulation testing. In addition to companies that are vertically integrated manufacturers such as Intel and Samsung, most companies only participate in some of these stages.
Just like having a blueprint before starting to build a house, in the chip industry chain, chip design companies are also at the top of the industry chain, and they are generally leaders who directly deal with customers.
Generally, different types of chips have different participating companies, and in the hottest GPU design field, the competition pattern is mainly reflected as one super and two strong companies. The super company is Nvidia, which accounts for more than 80% of the market share, and the two strong companies are mainly AMD and Intel, whose main source of income is CPUs, with a small percentage coming from GPUs, sharing the remaining market share.
Next is chip manufacturing. The technical level of chip manufacturing is very high and is considered the ceiling of modern industry. So which companies have the most advanced chip manufacturing technology? It's mainly TSMC and Samsung.
Taking the fourth quarter of 2022 as an example, TSMC accounted for nearly 60% of the market share of chip foundry manufacturing during that period, and Samsung's market share was about 16%.
Lastly, the chip encapsulation and testing link. Compared with the previous two stages, chip encapsulation and testing are relatively low-end, and it involves very few companies listed on the Hong Kong and US stock exchanges.
In addition to these three main industry chain links, there are many auxiliary links in the entire chip industry chain, and many listed companies have also emerged in these links.
For example, chip design companies often need to use specialized design software to design. The companies providing design software mainly include Cadence (CDNS) and Synopsys (SNPS).
Chip manufacturing companies need to purchase manufacturing equipment and raw materials to start production, and the entire manufacturing chain is very long, involving even more listed companies. For example, ASML Holding from the Netherlands provides lithography equipment, which is a core equipment in chip manufacturing and currently has a market capitalization of tens of billions of dollars.
3. Characteristics of the chip industry
Just as different people have different personalities, each industry has its own characteristics, and there may be different investment strategies based on specific characteristics. For example, the characteristics of the Internet are that the strong get stronger, and the top companies need to be given special attention, while the characteristics of cyclical industries are that they are prone to cyclical fluctuations, so it is important to observe the business climate of the industry. For growth industries, a long-term perspective can be maintained.
So what are the characteristics of the chip industry? It has both long-term growth logic and cyclical properties, and can be said to have both growth and cyclicality.
In terms of growth, as previously mentioned, the average growth rate of the entire chip industry is 7%, while the growth rate of the GPU subdivision industry exceeds 30%. This is because emerging industries such as cloud computing, blockchain, intelligent driving, and AI all rely heavily on GPUs.
In terms of cyclical properties, the historical market scale changes of the chip industry have the characteristic of high-speed growth for several years and then slower growth or even decline until the appearance of the next turning point.
The core reason why there are cyclical fluctuations in the chip industry is the mismatch between chip demand and production capacity supply. When chip demand rises and companies such as Nvidia receive orders to go to TSMC and other chip foundries to place orders for manufacturing, TSMC's production capacity may not be enough, so it needs to build a new factory and expand production capacity.
The cycle of TSMC's factory construction and production capacity expansion may last for two to three years, because it takes time to build new production lines and purchase equipment, such as one of the most core equipment, lithography equipment, which may take a long time to arrive.
After two to three years, TSMC's production capacity will be increased. At this time, the demand for the chip industry may have declined, and the industry enters a downward cycle, with excess production capacity and inventory backlog. Companies like Nvidia can only handle inventory by offering discounts until the inventory is gradually cleared, the industry demand slowly recovers, and the next upturn cycle is ushered in.
4. Investment logic of the chip industry
Based on an understanding of the characteristics of the chip industry, let's take a look at the investment logic of the chip industry. Different investment strategies are definitely used for its cyclical and growth characteristics.
First of all, for the cyclical nature of the chip industry, the focus of the investment logic is to determine the position of the cycle. If the judgment is wrong, it may buy at the bottom of the mountain and sell at the foot of the mountain.
To judge the position of the chip industry cycle, we can combine inventory turnover and the historical time cycle of the industry to make a judgment.
On the one hand, as we mentioned earlier, when the industry enters a downward cycle, oversupply will occur, inventory will increase, and inventory turnover ratio (cost of goods sold/average inventory) will also decrease. Therefore, this is a leading indicator that will respond more sensitively to cycle changes than revenue and net profit data.
On the other hand, according to historical data statistics, the average upward cycle of the chip industry is about 27 months, and the average downward cycle is about 14 months.
Therefore, if the chip industry has continued to prosper for more than two years, and the inventory turnover rate of the main companies in the industry has slowed down, or even turned from rising to falling, this may be a signal that the cycle is turning at the top, and investors may need to be cautious.
On the contrary, if the industry has been in a downward cycle for more than a year, and the inventory turnover rate of the main companies in the industry has slowed down, or even started to rebound, this may be a signal that the cycle has bottomed out, and investors may need to pay close attention.
Let's take NVIDIA as an example to see the changes in inventory turnover and stock prices.
During the period from July 2018 to October 2019, NVIDIA's inventory turnover rate began to fall from a rise, and its stock price also performed flatly.
During the period from January 2020 to October 2021, the epidemic stimulated the surge in demand for online office and cloud computing. The demand for chips increased sharply, and NVIDIA's inventory turnover rate began to soar, and its stock price also rose.
Of course, historical data will not simply repeat, and sometimes some demand outbreaks may change the operation of the original cycle. For example, the semiconductor chip industry is actually still in a downward cycle, but stimulated by artificial intelligence, the demand for computing chips has surged, and the market may judge that the chip industry is also approaching the bottom, and the industry is expected to improve. Chip stocks have generally risen well.
If the core of cyclical investment strategy is timing, then the investment strategy based on the growth of the chip industry is to look at its long-term value.
Many people may not have good timing, but they are optimistic about the long-term development of the chip industry and may adopt a long-term investment approach. However, if you are unlucky and buy at the top of the cycle, you may also experience a long waiting time to get out.
Some people who are not good at picking stocks can also pay attention to ETFs in the semiconductor chip industry. ETFs can generally achieve average industry performance. Most people cannot outperform the average performance, so considering ETFs may be a more suitable choice.
Finally, to summarize,
From the perspective of overall development, the chip industry is an industry with a very large market space and maintains a higher average growth rate.
From the perspective of industry classification and industry chain, chips can be divided into eight types from the application point of view. Logic chips represented by GPUs and CPUs account for the largest share, and the GPU subdivision segment has a higher expected growth rate. The industrial chain segments of the chip industry mainly include design, manufacturing, and packaging testing, while there are also many auxiliary segments, which have given birth to many listed companies.
From the perspective of industry characteristics, the chip industry, especially the GPU industry, has both high growth and high cyclical fluctuations, and the main reason for its cyclic fluctuations is the mismatch between demand and production capacity.
From the perspective of investment logic, the core of the investment strategy for the cyclical nature of the chip industry is to determine the position of the cycle. We can track and judge it through the indicator of inventory turnover rate combined with time cycle. For the growth of the chip industry, we can maintain a long-term focus.