To see a representative growth in the pharmaceutical sector, such as Bayer, I believe that we need to implement some form of interest rate reduction. So, let me show you one or two more slides. This one is about the market size of generative AI compared to the projected size of the AI semiconductor market. Whether this is right or wrong is not important; it's just what some institutions have presented. On the left, you can see that the generative AI market is growing tremendously. As you can see, it grows fivefold over five years. On the right, looking at semiconductors, the growth is only around 20 to 30%. The darker section at the bottom represents AI semiconductors, while the rest includes all system semiconductors. So, the semiconductor market is only growing by about 20 to 30%, and all of that growth is coming from AI semiconductors.

What this means is that, from our perspective, on one hand, big tech companies are currently investing in AI services without making any money. People are asking when they will actually start earning. In reality, it seems that the semiconductor companies are the ones making money. However, this could lead to a completely different narrative, right? There are two perspectives to consider. 

First, whether it’s Samsung, SK, or Intel, the truth is that all companies face a stagnation if they remain idle. In fact, if you take out the AI semiconductors from the graph, the dark blue part, you will see that the market is either stagnant or declining. Ah, I see. This means that if we exclude AI semiconductors, the semiconductor market is either stagnant or has already started to shrink. Therefore, if companies in the AI semiconductor sector do not grow, the moment they stagnate, they are already facing a decline, and their stock prices begin to tumble. Ultimately, this relates to the future of semiconductors.

The growth of AI semiconductors is absolutely limited when viewed in terms of potion. As you can see, the growth rate of the semiconductor market compared to the entire AI market is low. However, the growth that does occur is entirely within AI semiconductors. This means that AI semiconductors become the leverage for that growth. Furthermore, AI semiconductors play a crucial role in generating profits and enabling growth between memory and non-memory sectors. If I summarize, right now, HBM is the answer. There is a famous saying: At that time, it was right, but now it is wrong. Currently, HBM is the answer, but as shown in the graph earlier, the biggest bottleneck hindering the actual development speed of ARI and GPUs is the memory issue and the bandwidth problem. The company that solves this issue needs to emerge from South Korea. We, as a nation and as citizens, need to invest in this. Hynix.

Buy Samsung Electronics stock, but we don't know how it will turn out. There's news that Samsung Electronics and Hynix have captured more aliens, and if it’s said that they are preparing for a transition, right? This is, of course, something we need to approach more technically, but there’s a mix of anticipation and fear, and there’s also an exciting aspect to it. This could indeed be somewhat related to that era of the unfolding wave. Yes, yes, but why do experts, many of whom are musicians I know, refer to the unfolding wave as artistic? Jazz has a lot of unfolding waves, but modern music uses a lot of harmony. Anyway, we are moving toward that kind of world. However, the big change here is this: let me talk a little more about semiconductors. As you can see from the graph, among the types of semiconductors, the critical semiconductors...

This is a market outlook. CPU, GPU, and FPGA refer to programmable semiconductors. It means semiconductors that are designed and manufactured. Looking at the overall market, everything is developing. However, among these, the GPU has a relatively significant impact. It is starting to surpass the CPU. In fact, the FPGA market is also expected to grow significantly. Among these markets, we need to pay attention to the fact that, as I mentioned earlier, semiconductors can be broadly categorized into AI semiconductors. There are learning semiconductors, application semiconductors, and memory. I already talked about memory. Currently, only Google and NVIDIA can produce learning semiconductors, but AMD has now entered the fray. However, many people misunderstand that this is a semiconductor issue. Why would Google or AMD lack skill? It's not about the semiconductors themselves.

NVIDIA is not a semiconductor company; it is a software company. Jensen Huang has created an ecosystem. To put it simply, there is software called CUDA and an architecture that goes with it. This demonstrates overwhelming capabilities. So, we refer to the hardware as silicon, but just having silicon semiconductors alone cannot topple Jensen's stronghold. There needs to be a complete software architecture and platform, including CUDA, to create a system like NVIDIA. This is something that even Google, AMD, and Intel have not been able to achieve. 

So, the important question is: when do we need these components, including CUDA? The most critical contribution comes during the learning phase, when artificial intelligence is studying. Therefore, for the foreseeable future, the semiconductor that is used for learning and studying will likely be dominated by NVIDIA. It will be difficult for others to compete, and even if they try to replace it, they won't be able to match the cost-effectiveness.

It is not easy to come out or compete. So, where are all the other semiconductor companies focusing? It's inference, inference is easy. Inference is like taking an already learned artificial intelligence; now you don't think. No, you don't learn anymore. Just execute. When asked a question, just answer. If asked to draw something, just draw it. You don't study anymore; you've already graduated from college. It's the same as the dual-use. Our kids are studying in the U.S. now, and it costs so much money to study. So, when studying, it only costs money. But now that they've graduated, it's time to earn money with inference-type semiconductors. You can stop studying now. Yes, that's right. However, it's hard to know how meaningful this graph is. Yes, yes, yes. But if you look in the future, the inference semiconductor market is expected to be about twice the size of the learning semiconductor market. Yes, then inference-type semiconductors.

The market is indeed facing a downturn. Of course, this has been happening since the late 2020s to the early 2030s. So, who is making semiconductors? AMD is producing them, and Samsung is also in the game, along with others like Tenst and Jim Keller. As I mentioned earlier, semiconductors could be seen as a bubble. In some sense, it may be a bubble, but has that bubble burst? No. From the perspective of a high cycle, the expectations are just returning to normal levels. In fact, you all had expectations that were too high too quickly. Normal expectations should have been at a lower level. This is the potential that the market holds. If we had only expected it to be at that lower level, what would have happened? Investments should have been made at that time. Would you like to take a look at the graph? Then, let's see what the market holds.

The potential here is what we call expectation. The endpoint is here, not over there. So when should the investment have been made? The investment can be made when expectations are low. If the investment is made here, the expectation can actually double what it can produce. Therefore, the bubble needs to deflate to this level and reach here. This is the level of expectation. So what will happen with semiconductors? Well, AI software is the same, but currently, with Hynix and Samsung—especially Hynix—people are complaining. You are buying stocks, but they are going down. Oh, it’s hitting your heart. So, does that mean it’s a failure? No, it will go here. However, our retail investors can't wait until it reaches this point, so they sell everything here and end up failing again.

So, what I'm saying is that this time of disillusionment is like this. I bought Saltlux stocks here, but when I came here, I couldn't sell them. It's not possible now. I need to report to the person in charge. Oh, I have a call coming in from the company. So, ultimately, if you ask what perspective investors should have, it's about having expectations regarding the level. If you could show the graph one more time, it would illustrate the expectations. So, what kind of expectations am I holding? Those expectations are insights. I believe that Hynix semiconductors will also rise again. I firmly believe that Samsung and Hynix can achieve this for our nation. However, if you're expecting them to have risen to this point, Hynix has actually fallen to this level. So, the question is whether you can hold on or not.

I believe it's going to be a choice for everyone. However, as I mentioned earlier, how long will it take to get from here to there? Five years. Within those five years, can our companies overcome the structural and technological limitations that HBM currently has? I can't say for sure. That's the key point. If they can do it, then this graph will rise significantly. Why? Because we will be the only ones capable of doing it. Memory is important, but it's not the only thing. For example, it's just like what I mentioned earlier about making money by selling picks and axes. The ones making the picks and axes are companies like Diana and AMD. But do you know what South Korea is making? We're making axe handles. Without axe handles, we can't keep striking with the axes or picks. It's like if a millstone doesn't have a handle; it becomes useless.

It's unbelievable that we can't use a millstone. It's absurd. Ah, absurd! You must have watched a lot of movies. Well, this is currently happening in South Korea. But the possibility that the Chinese could make something absurd is real. That's truly ridiculous. There are many things to consider. So, whether I say to invest or not, whether there's a bubble or not, I don't think that's the important part. In the world of innovative technology, having bubbles is completely natural. During bubble periods, money gathers, which leads to the creation of innovative companies. If someone claims that they will gather 100 trillion won because of a bubble, then that person must be crazy. In such a case, one might think, Wow, it looks like 100 trillion won is needed. And even after calling for 100 trillion won right now...

Last time we were here, we were gathering 2.3 trillion won, and that money is coming together. Then, with that money, we can start changing the world, right? Thinking back, investing in OpenAI is something to consider. But you know, in the 90s, Jeff Bezos certainly did that, and those who were called crazy, including Elon Musk, are changing the world. Now we are in the era of Sam Altman. Recently, I saw that Sam Altman, who is the CTO of OpenAI, I think? I forgot her name, but she has resigned. Almost no one who was a co-founder is left; everyone has left. Originally, it started as a non-profit organization, but now it is leaning more towards profit-making, and Sam Altman definitely holds a strong position.

This may be a bit of a tangent, but I am somewhat worried. I'm concerned that in the pursuit of profit, technology might advance to the point where it creates something that could truly replace humanity, like Sam Altman. It's frightening. Did you know that Oppenheimer's birthday is the same? Oh, he's Jewish, right? Yes, yes, he's Jewish. I mentioned that one of his friends is Jewish too. Jewish people, especially, tend to be quite eccentric. Today, we're using some technical jargon here. If we think about the global eccentrics we know, as you mentioned earlier, who were they? Oppenheimer is Jewish, and Einstein is Jewish, and yes, and then there's Sethman, who is also Jewish, and Mark Zuckerberg from Facebook is Jewish too. I don't know why that is, but anyway, I'm considered somewhat of an expert domestically, and surpassing humanity is not an easy task.

So, I guess I wasn't expecting much. I brought it up to that point. Whenever I talk about it, it almost all comes out. It's almost like generative AI. Yes, we refer to that as AGI. Ah, officially, it's General Artificial Intelligence. So, on the left side, there's just a set, and on the right side, it's like that. People are now afraid of such things. This is what we call AGI. AGI stands for Artificial General Intelligence. In Korean, we could simply call it general artificial intelligence. Originally, it would be correct to say artificial general intelligence. So, simply put, it's like an intelligence that can do anything, just like a human. It's appropriate to view it that way. It can create anything, or rather, just at the level of a human. Understood, understood, at the human level. So, I thought about it. Then, what would it take to become human-level artificial intelligence?

해야 되냐? 우리는 모든 사람이 지능을 가져오기 때문에 그 지능이 얼마나 위대하고 얼마나 훌륭한지 몰라요. 그래서 인간이 하고 있는 놀라운 지능은 자율적 학습, 개념 일반화, 추론과 이해, 과학적 사고, 문제 해결 능력, 의사소통, 자각, 다중인지, 윤리, 사회성. 무엇보다 중요한 건 반성과 성찰이에요. 예를 들어, 오늘 들어가면 대표님이 내가 왜 그랬을까?라고 생각하실 수 있죠. 그리고 개성, 친밀감, 신체 지각이 아니라, 어쩌면 저희가 단순한 지능, 즉 읽고 답변하는 것을 뛰어넘는 어마어마한 지능을 가지고 있다는 거예요. 그걸 만약에 제가 레벨을 나눠서 설명하자면, 이게 대표님 레벨이에요. 휴먼 레벨, 저도 할 수 있는 거죠. 누구나 할 수 있는 거예요. 저희 어머니도 할 수 있고, 누구나 할 수 있는 사람이라면 할 수 있는 거죠. 그런데 지금 우리 레벨이 어느 정도인지 생각해봐야 해요.

Artificial intelligence, the best that artificial intelligence can achieve, is at a certain level. However, there are areas where it has almost caught up—especially in communication, natural language processing, emotional intelligence, and multimodal capabilities. It has made significant progress in these areas. But when it comes to other aspects, such as reflection and introspection, it's not easy. Ultimately, all these things require breaking through and surpassing certain limits. So, when you ask me how long it will take, I honestly don't know. It could be tomorrow, or it might never happen. However, from the perspective of researchers, experts, and executives like myself, it seems likely that it will happen after both the representative and I are no longer in this world, so we probably don't need to worry about it. It resembles a time depicted in movies, right? Well, we probably won't live for another hundred years. But anyway, that's the situation.

I don't know about that either, but I thought that in order for it to go well, we would need to implement a rate hike. I believe that it could give us another chance to jump higher.

GPU보다 더 큰 시장이 열립니다...외계인 잡아서 갈아넣는 이런 회사에 투자하세요 / 이경일 대표 (3부)

"지식과 자산의 복리효과" 그 길을 복길이 여러분과 함께 걷습니다.
( 업로드 : 오전 11시45분, 오후 4시, 오후 7시 )
제휴 문의 : ystreet2021@gmail.com

인터넷 언론사 제호 : 와이스트릿 / 등록번호 : 서울 아53714 / 등록일자 : 2021.05.24
주소 : 서울시 영등포구 국제금융로8길 27-9 여의도 동북빌딩 910호
대표번호 : 02-6949-1701 / 팩스 : 02-6949-1702
이메일 : ystreet2021@gmail.com
발행인 : 이대호
Copyright © 와이스트릿  YSTREET ALL RIGHTS RESERVED. 와이스트릿의 모든 콘텐츠는 저작권법의 보호를 받으며, 무단 전재·복사·배포 등을 금지합니다.
