Transcription
Hello everyone, welcome back to the second segment of "Stock Market C-View." We continue with Shi Sheng and James. In the previous segment, we discussed the situation with the China Securities Regulatory Commission. This time, we will talk about finance and technology topics. Huawei has recently made some new breakthroughs in chip manufacturing, which the market is also paying close attention to. It's called the "Tao (τ) Law." Simply put, in the future, chip performance will no longer be solely improved by making transistors smaller, but by shortening signal transmission time. It is reported that this method will elevate China's high-end chip technology to the 1.4-nanometer level within five years.
First, let me ask James. When this news broke, the semiconductor sectors in A-shares, including Hong Kong's semiconductor sector, all saw significant increases. How do you view this? And is it still a good time to invest and pay attention, or is it purely a short-term speculative play?
From a technical perspective, it's truly profound. I think those in the market who deeply understand it or grasp what it's saying likely require some professional talent to know. But we can look at one thing: the media hype surrounding it is very strong. This seems to be an attempt to create a whole trend around "Deepseek Moment" in early 2025. Because whether you look on Twitter, Facebook, or Threads, as soon as someone says, "This won't work," or "They're trying to run before they can walk," someone immediately launches into a lengthy rebuttal, saying, "Do you understand?" "Do you know what you're talking about?" "This is an advanced process." Honestly, as someone who isn't particularly familiar with this industry, after reading it, I couldn't really distinguish who was making a strong argument or whether the counterarguments were solid.
But I can think of one thing. Let's look at HBM (High Bandwidth Memory), which is a more familiar element that has been heavily speculated on recently. With HBM, as it progresses, there's a die, which is the smallest silicon chip. Eventually, these silicon chips need to be stacked up. The more layers you stack, the greater the probability of errors occurring in between. And if an error occurs in any layer, it will affect the performance of the entire HBM Stack. So, the ultimate challenge they need to overcome is yield and mass production. These are actually challenges that ordinary advanced process chips need to face.
Currently, we see that what they call "logic stacking" allows for a very high number of stacked layers. This means the length of the chip doesn't need to change much, and its size doesn't need to change much. I think the market challenge they ultimately need to address is, after achieving yield and mass production, and after production capacity ramps up, what will the actual production capacity be? I think this is a significant challenge.
Of course, you can look at positive and negative arguments. If we look back to 2023, the positive side is that the market previously underestimated China's ability to develop solar power, electric vehicles, and shipbuilding. Now, almost 50% to 70% of the global supply chains for these industries are concentrated in China. If we look at the negative side, perhaps at the end of 2023, Peking University and Tsinghua University proposed a new EUV concept. At that time, the idea was to bypass US EUV sanctions by using this EUV to create a Chinese-made exposure light source. However, there hasn't been much follow-up on this so far.
Or, previously, Deepseek was said to be a very successful app, but it also stated that it wanted to replace or supplement computing power with algorithms. This has not yet been validated. So, I think the logic here is that whether it's successful or not, they themselves will have to see by 2030. However, the resources and investment driven by policy inclination in the meantime are, in my opinion, a very long-term and sustainable investment. In Hong Kong stocks or A-shares, it's a major concept to invest in. That's why when this news broke, I think the market didn't have enough time to fully understand the authenticity of this concept before they poured money into the sectors and stocks they believed would benefit, including SMIC and Hua Hong. There are also other targets in A-shares.
We believe that since the policy stimulus in September 2024, the competition between China and the US in computing power, data centers, and chips has led these sectors, especially the largest ones that receive the most policy and resource倾斜 (inclination/bias), to benefit in the long term. With this new concept emerging, we view it in the same way.
So, do you think the overall sector should continue to be viewed positively?
Yes. Shi Sheng, have you looked into this Tao (τ) Law?
Of course, I understand even less. I can only see that the Chinese government, or China as a whole, is determined to take a different path in the chip competition. Previously, they followed others. When others achieved less than 10 nanometers, they followed. When others achieved 2 or 3 nanometers, they followed. They all wanted to obtain this technology through lithography machines. But this approach is essentially following others, making it difficult to surpass them. It's usually about overtaking on a curve. Now, they are trying to surpass from a different dimension. The mainland calls it "dimension reduction" to attack you. It's another universe, using a completely different method. If they can truly achieve this, then it will be their turn to lead, and the West will have to learn from them.
Of course, they have wanted to do this for a long time. Previously, there were different products that were ridiculed, changing their brand but actually copying from elsewhere. But I believe that even during this period, a lot of resources have been allocated. The state has allocated resources, large enterprises have allocated resources, and companies like Huawei have allocated a lot of resources to do this. The fact that they are announcing it so publicly this time suggests that the chances of success are likely not small. Even if what Deepseek claims is not yet achieved, after Deepseek's emergence, the world has begun to look at China's AI development capabilities with admiration. In reality, besides the US and China, no other country has the ability to participate in this competition in terms of AI. China may have some gaps, but it has established a different development model.
NVIDIA originally thought that the H200 would have a chance to be sold in China, but they really didn't buy it. At first, they just wanted the US government to allow it. The US government approved it, but China still didn't buy it. They couldn't even get the H20 before. If they didn't have confidence, they wouldn't have this boldness. If they could buy the H200, I believe it would bring immediate benefits to many Chinese companies. Now, they dare to refuse you. I don't think it's empty talk. So, the chip landscape seems to be changing.
From an investment market perspective, in the last segment, we mentioned that mainland funds can no longer buy Hong Kong and US stocks, or there are fewer funds. Can the semiconductor sectors in A-shares or Hong Kong stocks be considered? Please analyze this for the audience.
I think the market is very anxious, or perhaps a bit worried. AI has been developing for so long. We really entered the market's view from the first quarter of 2023. It has developed for about three years now. You see very few applications that can truly generate profit, while those with significant revenue increases are everywhere. But the problem is that it doesn't seem to help businesses, or the application-side companies, make money, yet everyone is still investing a lot of capital.
Take the US as an example. Historically, there has been a lot of infrastructure development in the US, building various railways across the country. Adjusted for inflation in today's dollars, it cost a total of $550 billion and took 70 years. Building the entire US highway system cost $620 billion and took 50 years. But if we talk about AI, just the investment in data centers alone in the past 5 years has been $930 billion. This is the fastest infrastructure project in history. With such rapid growth, it has, in my opinion, become a "too big to fail" concept. That is, if it ultimately turns out that "this endeavor is not feasible," it would be equivalent to saying that the US highway system and the railway system, these two major infrastructure projects, are not feasible. I don't think any stakeholder or investor can bear the consequences of this.
But the real problem is that you see companies like Anthropic, whose usage has increased 40 times in a year. It has just started to generate profit. At the same time, you see other companies, such as Microsoft, telling all its engineers to stop using Anthropic because it's difficult to quantify, and they feel the token usage is too high. You see Uber's investment in tokens has increased tenfold in the past six months, but the output has only increased twofold. So, they are starting to feel that this calculation might not be right. However, at the same time, you see Amazon and Meta still encouraging people to use tokens as much as possible, or Toshiba also encouraging maximum token usage. So, the market is a bit hesitant now. What should be invested in within the AI industry?
The direction of investment is very clear: all hardware bottlenecks. And these bottlenecks are constantly being identified, including GPUs (Graphics Processing Units), then CPUs (Central Processing Units), XPUs (Customized AI Accelerators), then cooling, power supply, then grid-related communications and transmitters, then HBM, and after HBM, people will talk about quartz, silicon interposers, and then some passive components. All these bottlenecks in the AI hardware sector are being identified by more and more professional people. And these people include... I don't know if you've seen on Twitter, there's a goddess named Serenity now. She has very clearly broken down the entire AI-related industrial chains in Japan and Taiwan.
Micron's stock has risen so much. But if you calculate its forward P/E ratio based on market-estimated earnings for the next year, it has risen from 5 times to 7 times. It's still in the single digits. If you look at Samsung and SK Hynix's P/E ratios, they are between 6 and 7 times. If they can deliver corresponding earnings, they are not ridiculously expensive. But they have already risen so much. So, I think the market's view on the entire industrial chain is: first, the difficulty of finding bottlenecks is increasing because the required technical knowledge is increasing parabolically. Second, the most easily understood ones, and those that are known to bring benefits, like Micron, whose target price was doubled by UBS last night (26th) to $1625 (a slip of the tongue). This kind of market stimulus is easier to find in overseas markets. But if you look in Hong Kong and A-shares, what you mainly look at is policy inclination. How much resource can the policy give you to make mistakes, effectively subsidizing your trial-and-error costs? Then, as you expand your reach, your market share gradually increases. This is what we want to see.
If we really talk about applications, I think Alibaba is a good example. So far, we haven't seen applications that can significantly increase its bottom line. Instead, because of increased capital expenditure, the increase in its bottom line has basically become zero. But if we look at the logic of why the entire Chinese e-commerce sector was once praised by global capital, it's because they felt the Chinese market was too vast. If you have a business operation and logic that can control this market, then our money will not be stingy in investing in this market and related companies. I think Alibaba and Tencent are two of the best examples. But the current predicament is that we haven't yet seen these mainland-related companies strongly convince the market with their applications, saying how AI applications can increase their bottom line by how much. But this predicament is not unique to China; companies worldwide are facing this dilemma.
Understood. So, during this period, funds are speculating on semiconductors. However, some of our viewers have asked why the Hang Seng Index hasn't performed as well, hovering around 25,000 points. Should viewers continue to "speculate on stocks, not the market"? Let me ask Shi Sheng. The Hang Seng Index's rise is clearly slower than other markets. Are you disappointed?
I believe most people who buy Hong Kong stocks are disappointed. One of the reasons is, for example, the AI craze in the US is far stronger than in Hong Kong. Of course, the rise of stock markets in South Korea, Japan, and Taiwan is also heavily influenced by AI. When things are hyped up so much, anything related benefits. For example, when digging for gold, the shovels are valuable, and the handles of the shovels are valuable. Everyone invests. Then everything becomes expensive. What about Hong Kong? I don't think there's this craze. Even if people buy, they buy US stocks, Taiwanese stocks, Korean stocks. There isn't a company in Hong Kong... Of course, some newly listed companies with AI-related stocks have risen, but their total market capitalization is vastly different from those abroad.
However, I've noticed some recent phenomena. Is a counter-climax imminent? I see some AI gurus giving lectures at universities and being booed. When they praise AI, the audience boos. Some are even at a loss, not knowing how to end. AI is being questioned or belittled. Recently, even the Pope has spoken, saying, "You people working on AI, don't compare yourselves to God. What do you think you can do? You harm the natural order and human ethics." Society is starting to see some opposing forces. Does this reflect that some of the hype has been too excessive? So, I think it's appropriate to be a little cautious and restrain ourselves, because things that go to extremes tend to reverse.
James, you mentioned the Hang Seng Index. Will you lower your target for the Hang Seng Index this year?
I really don't have very high expectations. For Hong Kong stocks, we always play by the first rule, as you said, "speculate on stocks, not the market." Related AI concepts and some packaging concepts are spillover effects from Japanese, Korean, and US stocks to Hong Kong. Some traditional, perhaps second-tier, packaging companies are doing well. Some companies that make transmitters are also doing well.
Another aspect is yield. We always believe in some giant state-owned financial companies or telecommunications companies. The yield they provide can be sustained, or the dividend payout ratio can gradually increase. I think we can consider these when the yield becomes attractive. However, overall, the index is not particularly cheap. The expected P/E ratio for the year, at its lowest point perhaps two years ago, was around 7.8 times. Now it's fluctuating between 11.5 and 12 times. It has risen from 7.8 times to this high. If you say it's a value洼地 (low-lying area/undervalued spot), the persuasiveness starts to decrease.
Another point is its earnings. Indeed, due to the absence of any star AI stocks, the speed of earnings growth is not as strong as other indices in the market. The expected increase in earnings per share for the Hang Seng Index, the total increase in earnings per share, I think this range is around 3% to 10% this year. The expected earnings growth for Japanese and Korean stocks this year is 16% to 18%. US stocks, for the entire year, are expected to grow around 10%. However, in the just-completed quarter, the S&P 500's total earnings per share increased by 27%, due to the combined efforts of some giant companies and smaller companies. So, earnings and valuation are the two most important factors in determining whether the entire market can rally. I think, for now, in Hong Kong stocks, I'm more conservative.
Thank you very much for James's analysis. Time is almost up for today. We hope to invite James again next time to share your views with us. If our viewers like our program, please subscribe and leave comments. We will see you next week.