저자: CathieWood · 외국인
요약 보기
미국 경제가 금리 상승·장단기 금리차 역전·관세·정부 셧다운·전쟁 등을 극복하며 고성장할 것이라 주장. AI·로보틱스·에너지저장·블록체인 등 기술혁명으로 세계 GDP 성장률이 현재 3%에서 3~5년 내 7% 이상으로 상승 전망. 전기화로 석유 수요가 줄어 유가가 안정될 것이라 봄. AI 규제는 필요하나 위험만 강조하면 안 된다고 언급. 엔비디아는 AI 인프라 확대 수혜로 긍정 평가되며, 메모리·전력·네트워크 병목이 다음 투자 포인트라고 설명.
다듬은 글
Uh, Katie, I...I'll tell you what my summary of what you said is. The US economy is confidently overcoming rising interest rates, an inverted yield curve, tariffs, government shutdowns, and war. It continues to generate high growth rates, and you will see this continue in the years to come with the addition of what could be comparable to the industrial revolution—namely, AI.
Certainly. Well, and in fact we are drawing on the experience of the industrial revolution. Well, if you look back 400 years before the Industrial Revolution, Bret Winton, our chief futurist, did this work. Well, real GDP growth in the world was about 0.6%. And then, with the beginning of the industrial revolution, which was the last technological revolution, uh, we got to 3% growth in global GDP.
Oh, and that was a fivefold increase. Well, we think that over the next three to five years, we will see, on a regular basis, and perhaps on average , growth of, uh, over 7%. Well, and that's only two and a half times more, uh, despite the fact that we have much more technology developing today than we did then.
Back then, it was electricity, the telephone, and the industrial internal combustion engine. Now we have, uh, AI, which is important, robotics, energy storage, blockchain technology, and multi-omic sequencing in the life sciences. Well, so, uh, a lot more technologies are developing a lot faster. So I actually think 7% will look conservative in retrospect.
There are many other, uh, interesting notes. And here we have a closed straight, and, you know, fears about inflation, you know, are on everyone's lips. You believe it is on the verge of long-term collapse. Why? Yes. Uh, I think, uh, productivity growth, which is important, uh, is one of the reasons.
We also believe that oil prices, uh, will come down quite significantly, uh, as transportation moves to electric grids. Oh, and, uh, the power grid doesn't use oil. E , 57% of oil consumption is for transportation. Well, and that's why we believe that as transportation moves to networks, the demand for oil will decrease.
And there we say, and there is a graph that shows that every time we are in a crisis, the price of oil usually drops to $30-$35. Uh, we think it won't be a crisis, except maybe for, uh, oil companies and energy-dependent countries , uh, but we think it could happen again, but this time it will be stable.
Industrial economics . And, and, and here's the thing. The article talks about it, and it's not uncommon, uh, for companies to, you know, hold these kind of, you know, war games or something like that. The Pentagon does it, big companies do it. But in this particular case, these executives believe that a major incident is inevitable.
And what will they do the day after the disaster strikes? And, you know, you're in the very center of this industry. So , what do you think about these endless tales of doom and gloom? Yes, I think it's important to understand that with any new technology, risks are inevitable. Like, say, when we invented fire, it allowed us to cook food, reduce disease and all that, but it also led to houses burning down and things like that.
Not very good for Prometheus either. That's right. Yes. And I think, like the internet, we had the same thing, and I think there are risks with artificial intelligence, and it's important to talk about that . But if we're only talking about risks, I worry that the same thing will happen as with nuclear power, where you only see the negative, the bombs, instead of getting almost free and affordable energy.
And we need to see the good, not just the bad. So we need to talk about the positive aspects of artificial intelligence. I have a book at home about the first trains, which I remember going 50 miles per hour. And this was met with resistance, because people really believed that when a train passed by , the atoms in their bodies could explode.
So you're saying that AI governance is a useful first step that needs a clear standard. When you talk about a clear standard, should it be an industry standard or, you know, and who should set it: the industry or the government, because that's also like a tug-of-war zone right now. Look, I think this industry is led by a lot of incredibly smart people with very strong beliefs and, frankly, some animosity.
These are people who all worked together. Some of them don't like each other, and that's why they started their own companies. And so I think the fact that they all got together in the same room in the White House , discussed this, and signed the same piece of paper is an incredible first step, and I think it's important.
Now, should the government be involved ? I guess to some extent yes, because if you just get a bunch of industry leaders together in closed offices to talk, antitrust issues arise, so to some extent yes; But I actually think we need to accelerate the adoption of this technology for good, and I think getting leaders together in one room to discuss it is a positive thing.
So when Astra came out and solved a math problem that had been unsolved for maybe a hundred years, and oh my God, I don't have it at hand right now...one of these mathematicians, he was really pissed about it, you know? He...we showed on screen how he was worried about losing his job, and...and what I need most is to be understood.
We really do what we do, he really made a public statement. So, of course, this week, you know, artificial intelligence is solving a lot of old math problems , and mathematicians are just going crazy. People say we wo n't need them anymore. And I've read that you find it fascinating, but many say it's also destructive.
Yes. I mean, basically, I think if you look at the progress in mathematics in relation to AI, in 2024, a lot of AI labs have presented results of how they're just solving these exams , basically. A month ago, we saw Anthropic and OpenAI solve these age-old mathematical proofs. And this is a huge step forward in two years.
And now, 3 days ago, OpenAI says, "Hey, we actually solved 700 of them ." And so my point is that it's good that we're solving these problems. Mathematicians all signed this document saying that no, this is bad for mathematics. And their reasoning is essentially what your teacher told you: it's not about the solution, it's about the process of solving.
That's right. And these AI proofs don't show the train of thought, they don't actually lead to human progress in mathematics. And my opinion on this is this : you can read the proof and you can accelerate your own learning. I think that's good. We must solve the world's most important problems, and we must not slow down.
Let me ask you about AI models: open versus closed. You know, we saw Muse take off, huge adoption. Grok bot is feeling very good. Although Muse doesn't charge a fee yet. So, you know, is this a discussion...we see on screen that the developers of Frontier are making all the money, even though they're losing market share.
Mhm. Is it simply because they focus on the corporate sector? Is it because of this? Yes. I believe that there is more money to be made in the corporate sector now . And they simply have a higher margin because of the token fee. But I think we've already discussed this in past shows. Almost no one disputes that models are becoming a mass- market commodity.
And the biggest proof of this is Muse, Grok, Instinct. And Elon Musk runs one of the largest laboratories in the world. He has his own model. He could charge a token fee, but two days ago he tweeted: " You know what?" "If you communicate with Grok, we will find you the best model for this task." This means that it doesn't care what model you use.
It is important to him that you use his interface. That's where value will accumulate. OK. You just published a piece about what the industry needs next. These products can do more. But conversation remains the primary way to manage them. What do you think will happen next? What do you think the next step will be ?
The next big milestone, so to speak? Yes. I think AI was a toy at first, and then it became... Before we get to today's topic, if you want deeper market analysis, exclusive information, sponsorship. This will help support the channel and give you access to additional content about the market. Nvidia is approaching a milestone that once seemed nearly impossible.
of $6 trillion. And stocks are just below record highs after five straight strong sessions . And according to Raymond James analyst Simon Leopold, Nvidia may still have significant growth potential. His target price is $515, one of the highest on Wall Street. But the big news isn't just Nvidia's stock price.
And what is happening in the entire AI infrastructure market. Because demand for Nvidia chips continues to grow across various sectors of the economy. Hyperscalers spend huge amounts. New cloud companies are raising capital, and companies like SpaceX are preparing large-scale investments in AI infrastructure.
And even though the cost of borrowing remains high, demand for computing power has not slowed significantly. This is a key point. This is not the only company betting on AI. This is the development of an entire industry. And And sits right in the middle of this. There is also a serious misconception about AI trading.
Investors sometimes perceive the market as if there will be one winner and everyone else will lose, but the reality can be much more complicated. There may be open and advanced models, hyperscalers and specialized cloud providers, different types of processors and network architectures , different applications.
The market is expanding in many directions simultaneously. And Nvidia is positioned as one of the critical infrastructure providers that connects many of these pieces. But there is a problem. The bottleneck is no longer just companies' desire to get more GPUs. The physical infrastructure needed to build these systems is becoming increasingly difficult to provide.
Memory is one of the biggest limitations. AI systems require vast amounts of high-performance memory, and demand is growing faster than supply. This imbalance has already pushed memory prices up, and it's not just about that. Energy consumption becomes another critical constraint . Data centers require enormous amounts of electricity.
They need a network. They need cooling . They need storage and advanced components, and it all needs to work together seamlessly. So while investors continue to focus on demand for GPUs, the next stage of AI development may depend on everything surrounding these GPUs. This is where the concept of connectivity becomes extremely important .
The first phase of the AI boom was dominated by computation. Then memory became a big part of this story. The industry is now increasingly focused on transferring information between all of these components quickly enough to keep expensive equipment from going down. Think about what this means. You can create enormous computing power.
You can install a huge amount of storage, but if those systems can't communicate effectively, you're not getting the economic benefit from the infrastructure. And as AI moves to more complex reasoning and agent-based tasks, the architecture becomes even more complex. Different types of computers must constantly interact with each other.
This is another huge opportunity for networks, optical connections, switching, and related infrastructure. One analyst during the discussion noted how dramatically the market for specialized and alternative AI infrastructure has expanded. The AI market, once estimated at around $14 billion , could now reach around $235 billion .
And this shows how quickly expectations for AI infrastructure have changed. This also explains why the possibilities are not limited to Nvidia. Opportunities in the semiconductor industry are expanding. Compute, networking, memory, storage , connectivity, security, infrastructure software—all of these levels can benefit from implementing AI into everyday business processes.
And keep an eye on. The economics of AI are starting to move beyond the idea of simply buying more chips. The real question is what economic benefit companies can get from all this computing power. And this is where generative AI gets interesting. Instead of AI simply answering questions, imagine digital workers working side by side with humans, continuously.
They can process information , analyze data, write software, monitor systems , perform routine tasks, and work around the clock . If this happens on a large scale, the economic value of the AI cloud could become much greater than current software applications predict . But their implementation requires huge capital, and therefore capital formation may become one of the biggest challenges in the next phase of the AI boom.
Comments 0