저자: 미상 · 외국인
요약 보기
AI 모델 가격 경쟁 심화로 LLM 업체는 가치 획득이 어렵고, 가치는 하드웨어(반도체)에 집중. 하이퍼스케일러 현금흐름이 반도체로 흘러 최소 6~12개월 프리미엄 유지 전망. 반도체 주식은 언젠가 폭락할 수 있으니 2~3개월마다 손절선 상향 권고. 장기적으로는 로봇·자율주행·공급망 혁신 기업이 유망.
다듬은 글
Guys, AI has a huge problem right now, and investors are starting to notice. I don't know if you saw this, but Amazon held a bond sale this week to raise more money for AI spending, and it got an extremely cold reception from investors. Grock released a brand new AI model that was actually pretty good and cheap.
Meta did the same. Broadcom has signed a huge deal with Apple. But here's what smart investors are starting to notice. AI has a huge, terrible problem, and that is the price. I'm not just saying that. Getting clicks and views and creating content is exactly what OpenAI co-founder and CEO Sam Allman said today on CNBC, introducing his brand new AI model.
That's what he said. Customers, obviously, every enterprise is now thinking about the costs and the value they get in exchange for AI, and this is the first year that AI costs have become a big topic, and suddenly it's a very big topic. Everyone is asking what we can do to help reduce costs or increase value.
Now imagine spending hundreds of billions of dollars on a revolutionary technology that no one has ever seen before, and then customers just slam the door in your face because of the price. Imagine investing over $500 billion in computing to create this revolutionary technology, as OpenAI did, and seeing competitors release products with similar features at a lower price point day after day, week after week.
Guys, if AI was a T-shirt or car company, you would bankrupt those companies. But here's a mistake that investors start to make, and I actually see it in the comments sections of my videos when I talk about this topic. Investors are confused about where the value level of AI actually lies and where it is actually being created.
In today's video, I'm going to explain all of this to you . This way , you will have a clear idea of what to do with your semiconductor stocks like Broadcom, AMD, Micron, Nvidia that have gone up so much, or with the software stocks you bought because some YouTuber told you to buy them , and they haven't gone anywhere, or worse, they've gone down.
I'll discuss all of this in today's video . Let's look at this in more detail. So, as I said, artificial intelligence has a big problem, and the first problem is actually the cost. Large language modelers like OpenAI and Anthropic have to spend hundreds of billions of dollars to build their brand new large language model and their products.
Hyperscalers like Google, Microsoft, and Amazon also have to spend hundreds of billions of dollars, all together, on a product that cannot be sold at a premium price. And it's not just me saying this , or me coming to this realization. Here is the CEO of PaloAlto Network saying that the price needs to be 10 times cheaper for the product to be usable.
We need to see AI prices come down. I think today's pricing even makes it very difficult for enterprises to adopt the rate, and we don't need the friction caused by model companies in terms of pricing and deployment. I think 54% is a good start. I think we probably need one more step. But I think if we can get the token efficiency from where it was to about 1/5-1/8, that's a good indicator for the next 12 months.
And I expect that by next year we need to get to about 110. So AI models with a big programming language look like two gas stations on the corner, and the one with the lowest price per gallon wins. So what should an investor do? And more importantly, where can you make big money in the next few years, when an AI model with a big programming language will be like gasoline at the gas station?
So let's answer that first question. In the comments section, I was asked if they should sell their shares in semiconductor companies? Should they sell the semiconductor ETF they are in? Should they sell to Nvidia, Broadcom, Micron or AMD? And my answer to that question is: why do it at all? The money that AI companies and hyperscalers will beg, borrow, and steal to spend on AI goes straight to them.
Look at this chart from Bank of America . Hyperscalers funnel every ounce of cash flow they generate into the semiconductor industry. And this will continue until the end of this year and probably until 2027. Quitting now is like selling your gas tank before the automobile was invented. And that brings me to the next point.
Every investor watching this really needs to understand where the value level in AI comes from. It does not come from OpenAI. It does not come from anthropogenic, coding, genetic assignment, or anything else. In fact, it is at the hardware level. Nvidia makes a product that almost no one else in the world can make.
So they print money. They can charge a lot of money for it. Same with Micron. Same with AMD. Same with Broadcom. And now I want you to flip this script. And once you do, you'll understand exactly why the level of value doesn't exist at these big language modeling companies. Just look at Grock this week.
I mean, it was about a day ago. This is obviously part of SpaceX now. They released a model that many in the industry consider competitive with many of the large language modeling companies we all know. But Meta also did the same thing . Chinese open source models appear every day, every week. Microsoft is replacing its Frontier models with open source models or ones they developed in- house.
You can't do this at the hardware level. You can't do that with Micron memory. You can't just find a company that sells GPUs on par with Nvidia. But at the level of large language models, you can find dozens of alternatives, and new ones appear every week. That's why Sam Allman has a pricing problem.
That's why PaloAlto Network CEO believes the cost should drop 10x or more. Big language models are selling dirt to gold diggers when all they need are shovels. So what do you as an investor do now? Where will the big money be made? First, let's understand that value creation happens at the hardware level, semiconductors a bit at the data center level.
And look, this will continue and they will be able to charge a premium price for at least another 6-12 months. Here's what I want you to do. Pick some levels for your Nvidia or Broadcom or AMD or Micron stocks that you can't survive a stock drop at, and just review that level upwards every two to three months.
That's all you need to do. Semiconductor stocks will crash someday. Trust me, the industry has always done it this way. Anyone who thinks things will be different this time is crazy. Just like the price of shovels and mining equipment after the gold rush ended. The same thing will happen to semiconductor stocks.
Just have a plan, go out there, and feel the joy that you're not left with an empty bag of money in semiconductor stocks and you don't have to wait years and years and years to get that money back. Now let's move on to a more important issue. Where can you make big money outside of semiconductor stocks ?
Listen, they won't be on the level of models. These companies could just as easily sell gasoline, toilet paper, and plain white T-shirts, as Gunness does. There will be no Louis Vuitton AI and no Walmart AI, just as there will be no Rolex gasoline and no Timex gasoline. Everything will be the same. So, where is value created in this environment?
It is actually created through innovation and implementation. Robot manufacturers, self-driving cars, companies that are reorganizing their ancient supply chain or customer service. These companies will win. Just like airlines, every airline, think about it, every airline uses gasoline for airplanes.
They all fly the same planes. But those who can sell first class seats are the ones who make the money these days. There is a race to the lowest price with major language models and their pricing. And it's not just me saying this, okay? Hardware or semiconductors are where you need to be an investor because they are so hard to replicate and there are only a few companies that can actually produce them.
And also innovative companies that create new, unprecedented products. They will be huge winners in the field of artificial intelligence. That's all for today. I hope you enjoyed today's video. Come back again on the main channel tomorrow. Good luck with your investments.
Comments 0