저자: 미상 · 외국인
요약 보기
출연자는 2013~2020 기간 미국 플로리다에서 매크로 중심 해외 투자 체계를 구축했다고 설명하며, 2018년부터 엔비디아를 보유해 2023년 비중을 늘렸다고 밝힌다. AI 수요와 Rubin 등 차세대 GPU로 토큰 비용이 낮아져 추론 수요가 확대되고 GPU 수명도 길어져 마이클 버리의 감가상각 비판은 틀렸다고 주장한다. 엔비디아 선행 PER이 약 17배로 낮아졌고, 2027년 매출 성장 가이던스(약 70%)를 감안하면 2027년 주가 기준 PER 12배 수준이 되어 50% 이상 상승 여지가 있다고 본다. 다만 AI 기술 변화가 기존 가치를 훼손할 수 있어 밸류에이션 안전마진이 필요하다고 언급한다.
다듬은 글
Hey, clear up a doubt for me because, okay, I understood that you tried to stay in Brazil, but you couldn't because you couldn't work and you felt underutilized. That is. But someone who experiences the market as intensely as you did, doesn't stop trading, right? How did you manage to stay afloat during that period?
In 2013 you left GAP and until 2021, that is, until 2020, it was 7 years . Yours. Very simple. I created a structure in the United States, in Florida, where I stayed, that had everything. He had independent research, he had Bloomberg, he had an absolutely complete office for business analysis. I had several, nowadays there are websites that do the whole valuation part.
So, somehow I developed, I became 100% international. That's when I said, " No, I don't want Brazilian assets anymore, I want to truly internationalize." And obviously, in the first few years I stumbled a lot, but then things started to fall into place. I had quite a bit of capital and I had spectacular results for myself, you know?
But it was more of a bottom -up thesis. I was starting to develop this idea that the macro is more important than the micro, right? And when the BTG era arrived , I already had that pretty solidified in my head, I was already operating that way, you know? So, I was already creating a macro scenario, I was already developing it, and in reality, this demographic story was just another adjustment to that scenario, right?
I, for example, have been investing in Nvidia since 2018. Wow. Yes, I mean, it's just that he didn't invest in the rest, he didn't make the connection between the issue of productivity and the issue of demography. But what was your thesis about Nvidia in 2018? No, well, I thought Nvidia was, I mean, in the case of AI, I thought it was a good possibility.
Obviously, there was little talk about AI. I believe that moment occurred in 2023. And I thought that this was one of the companies that had a great opportunity to take advantage of this technological advancement. At that time they, they met; Nvidia had a rather curious situation, because when they started, they made GPUs that had the clear objective of serving the gaming market, right?
Because, if the processing capacity were slow, the games would be frame by frame. Then came the era of crypto, and in a way, crypto needed superlative processing power and started using GPUs, graphics cards, right? So, when AI arrived, when this topic of AI began to be discussed , the need for computing power was obvious, right?
You had, you were going to generate intelligence, you were going to perform data manipulations and interactions , and that was going to create that intelligence, right? And it was very clear that more computing power would be needed . And Nvidia kept evolving, showing, for example, more advanced GPUs each year.
It started with the H100 series, then the H200, then you had the Blackwell, and today you have the Rubin, which is a completely different capacity from the previous series. Today they achieve 30 times more computing capacity for the same megawatt of electricity consumption. They manage to generate 30 times more tokens than in the past.
In other words, you're clearly paying more for Nvidia, for the Rubin Vera GPU today, but you do have the possibility of reducing the tokenization cost. In other words, an example I heard today is very interesting. Imagine it's almost like you're going to fill up your car with fuel and you're going to spend 200 euros filling your tank.
And with that you would travel 500 kilometers. With the Rubin Vera you'll spend 50% more, because it costs 50% more, so you'd spend 300 , but you won't travel 500 kilometers, you'll travel 10,000 or 15,000 kilometers . That's the truth. So, this story makes sense, right? So, something interesting today, right?
You see that Michael Burry, who raised this issue in late 2025, right? He explained that these investments could not continue in this way because depreciation, that is, the inability to meet the computational need required at that time and also energy inefficiency, which is an effective cost, would cause the infrastructure of a data center to last 3 years, and that, in 3 years, he judged that in fact no return could be obtained in this story .
Look how curious, the H1 series, which is already celebrating its 9th anniversary today; When you look at GPU H rentals, it's increasing; that is, with inference, you don't need to have the latest technology. To make inferences, you need the state of the art to perform the training, right? So this has created an extension in the lifespan of these GPUs, and this, in a way, has generated an increase in the profitability of these data centers, of the people and companies that invested in this infrastructure, right?
But this is incredible, isn't it ?! Because this shows how wrong Michael Burry was. Yes. Michael Burry, for those who didn't make the connection , is the one who became famous for "The Big Short," which was even made into a movie. Unfortunately, after "The Big Short", he never got anything right again, did he ?
Although it took "The Big Short" quite a while to get it right, didn't it? He was a couple of weeks short and he would have run out of money, but out of curiosity I went to see what Nvidia's market value was. In 2018 it was worth approximately $81 billion . Exactly. Today it stands at around 5.3 trillion.
Yes. So, did you keep it from then on, or did you make some sales along the way? Do not sell. You certainly sold little, but you got to No, I always had Nvidia, but I would say that there, in the year 2023, I increased my position in Nvidia more relevantly . Yes, it was very good. Wow, you've been pruning it ever since, otherwise it would have turned into...No, no doubt about it.
Yes, there is no doubt. But hey, you know that today, if we really look at it, it's even difficult to say this, right? But today Nvidia's future price-to-earnings ratio is at 17 times. Even today I saw that it was falling by 2% . It's possible it's in the 16.8 times range, right? If we keep that same one, people, the translation so that it's understood is price over benefits, right?
The "forward" is the price on projected earnings, right? So, if we maintain the same price for 2027, given this company's expected profit growth and revenue growth, right? Just to give you an idea, in the guidance given in the second quarter, the market expected 45% revenue growth by 2027 and Jensen said: "No, it will be 70%, it won't be 100% only because there are limitations in the inputs needed for GPU production, etc." But the demand exists.
That was their guide, wasn't it? But, that is, if we keep the price constant in 2027, Nvidia's future price-to-earnings ratio will be around 12 times. And if we maintain the same price through 2028, it might end up between 9 and 10. So, this means the following: how are you going to have, how can you allow a company of this quality, with margins around 70%?
Obviously, margins are going to give way a little, uh, throughout 2028, but revenue growth of this order, for example, was 100% last year, will be 70% next year. How can you imagine that this is going to be worth it, or cheaper than companies that have no relevance, no comparison in relation to these metrics?
So, simply put , what do I consider fair? I think that nowadays, I believe that AI actually poses some risks because this technology can evolve in a somewhat more aggressive way and that could disrupt some things. Some things that may make a lot of sense today may not make as much sense in 2 or 3 years.
So I think there has to be a margin of safety in terms of valuation for all companies, right ? So it's no coincidence that the S&P is currently at 19 times. I think it's already on that path. But considering what Nvidia is, imagine this company being valued at 18 to 19 times. 18, 19 times. You can assume that Nvidia will have more than 50% upside by 2027 , which is a huge amount, you know?
Wow, that's crazy. And, interestingly enough , even with the valuation falling, the price didn't fall. This is a sign that, in fact, what happened was a superlative increase in the company's profits. That's the truth. If you liked this clip, you'll definitely like the full episode. So , look, click here to watch it in its entirety, or you can also click on this video and watch another clip from this episode. Until next time. Yeah.
Comments 0