저자: DanIves · 외국인
요약 보기
결론: 데이터센터 규제·정치 리스크로 인한 AI/빅테크 하락은 과장되었으며 하락은 매수 기회라는 낙관론이 중심이다. [정치 리스크] Dan Ives는 중간선거 이후 민주당 '블루 웨이브'가 현실화하면 AI·데이터센터에 대한 규제와 모라토리엄이 강화되어 기술주에 부정적일 것이라고 봤다. 다만 어떤 정치 수사가 실제 리스크인지는 투자자가 판단할 문제라고 선을 그었다. 규제로 데이터센터 건설이 늦어지면 혁신이 저해되고 중국에 뒤처진다는 논리도 덧붙였다. [실적 전망] 다음 주 하이퍼스케일러 실적이 시장을 안심시킬 것으로 기대했다. 메타 실적이 나오면 주가가 오를 것이라고 전망했으며, 최근 메타·스페이스X 반응은 오히려 매수 기회로 봤다. 뉴욕 택시기사의 비관론은 무시하라며 비관이 극단화되는 시점을 매수 신호로 해석했다. 현재 하락을 'SaaS 아포칼립스 2.0'으로 규정하고 탈출 버튼을 누를 때가 아니라고 주장했다. [Gene Munster 인터뷰] 메타 Muse 일일 사용자 1,000만 명과 저커버그의 회사 중심축 전환 발언, 테슬라 FSD 설치율(3년 전 약 15%→현재 70%), 젠슨 황과 머스크의 컴퓨팅 확장 계획을 근거로 AI는 아직 초기라고 강조했다. 개인화 봇과 에이전트는 일반 챗봇보다 토큰을 10~15배 더 쓰며, 피지컬 AI 가능성까지 더하면 과거 몇 년보다 큰 변화가 온다고 봤다. 지금의 불안은 시장이 기업의 확신을 받아들이지 못하는 간극에서 비롯되며, 결국 진실이 이긴다는 입장이다. [자본지출 우려] 시장은 내년 자본지출을 약 38% 증가로 보지만 실제는 70% 이상일 수 있다는 수치를 제시했다. 2028년 기준으로는 시장 예상 11%에 비해 실제 25~40%가 될 것으로 봤다. 이 경우 주식시장이 긍정적으로 반응할지는 확신하지 못했다. 진행자는 경쟁 구도에서 과잉 투자가 구조적으로 나타날 수 있다고 지적했고, Munster는 기업들이 이웃을 따라서가 아니라 확신에 따라 투자한다고 답했다. 틀리면 '종말의 시작'이 될 수 있다는 점도 인정했다. [엔비디아 유럽 이슈] 유럽의 AI 컴퓨팅 점유율은 약 5%, 미국은 약 75%로 제시됐다. 젠슨 황은 GTC 베를린에서 유럽에 서두를 것을 촉구했고, 'Vera Rubin DSX' 컨테이너형 데이터센터 설치 기간을 수개월에서 수일로 줄일 수 있다고 발표했다. 엔비디아는 지난 1년간 유럽 스타트업과 파트너에 100억 달러를 투자했다. Procore는 엔비디아 Omniverse 기반 디지털 트윈 기능을 클로즈드 베타로 선보였으며, 엔비디아도 자사 AI 팩토리 건설에 Procore를 사용한다. 디지털 트윈은 입력 데이터 품질에 의존한다는 점도 강조됐다. Procore의 DroneDeploy 인수(8억 4,500만 달러) 언급도 있었다. [로봇 오리 및 허깅페이스] Pollen Robotics의 Micro Duck(25cm, 399달러)은 허깅페이스 소유이며, 엔비디아가 허깅페이스를 약 130억 달러에 인수하기로 합의했다고 언급됐다. 거래는 2027년 이후에야 완료될 가능성이 크다는 보도가 인용됐다. 사전 주문은 2만 건을 넘었다는 보도가 있었다. 진행자는 디지털 트윈, 데이터센터 구축, 피지컬 AI 로봇, 개발자 커뮤니티 확보를 아우르는 엔비디아의 단계별 전략으로 정리했다. 엔비디아는 칩 판매를 넘어 AI 기반 세계 전체를 판매하려 한다는 결론이다. [리스크·조건] 정치적 블루 웨이브, 규제 강화, 자본지출 과잉, 실제 수익화(매출) 증거 부족, 컴퓨팅 비용 증가가 주요 리스크로 제기됐다. 낙관론의 전제는 빅테크의 확신과 실적 확인이다. [출연자 전망] Dan Ives는 기술주 하락을 매수 기회로 보며 다음 주 하이퍼스케일러 실적에 낙관적이다. Gene Munster는 AI 초기 단계라며 지금 주식을 사기 좋다고 보았다. 다만 Munster는 자본지출 증가율이 시장 기대를 크게 웃돌면 주가가 긍정적으로 반응하지 않을 수 있다고 우려했다.
다듬은 글
Dan, you mentioned the midterm elections. Is it just about data centers? After all, dozens of lawmakers spoke out yesterday after an Axios report about how AI agents are hacking into people's bank accounts. Most notably, Bernie Sanders says that AI development should be halted immediately. Is there a greater risk that something will happen, given that Democrats appear to be poised to take control of the House of Representatives and perhaps even the Senate ?
Well, I think the concerns about the midterm elections, and I think it's not been talked about as much because we're obviously talking about interest rates and 10-year bonds, but a lot of the concerns that I see are purely political in nature. Because if there is some kind of " blue wave": Democrats are negative about AI, negative about building data centers, like in Holcomb, and then the moratorium that we see.
Listen, this continues. This is a real threat. After all, the reality is that with every data center that is not built in the US, China wins. So, I think from an investor's perspective, you're going from state to state right now. If something isn't being built in one state, they move to another. But I think right now, from an investor's perspective, it's certainly a concern ahead of the midterm elections.
So if this is the " blue tsunami" that some Republicans have warned me about, do you think all these hyperscalers will open at significantly lower levels on November 4th? Listen, if there was a " blue wave" or a "blue tsunami," it would be negative for the tech sector, no doubt about it. Now the question is what is just high-sounding political rhetoric and what is reality.
And I think, ultimately, investors will figure it out themselves. But I think that's the main problem right now, because any regulation or slowdown in the construction of data centers is what stifles innovation, because the US has overtaken China in technology for the first time in 30 years . And that's a mystery right now, but I believe that the hyperscaler reports next week will reassure the market, showing that, in my opinion, we are the leaders.
I need to clarify. Maybe we'll take Meta and SpaceX. The indicators were quite good. This did not reassure anyone. Yes, and I think the reaction to Meta and SpaceX in the coming weeks and months will be seen more as a buying opportunity, given how the market has reacted. I think if Meta reports earnings next week, the stock will go up .
I just think it happened at a time when nervousness reigns supreme. And now, you know, even New York taxi drivers are pessimistic about the technology sector. We see a lot of nervousness, but I view this as “SaaS apocalypse 2.0” and not as a time to hit the “exit” button. Ignore New York taxi drivers.
When you become pessimistic—that's a signal to me that it's time to buy. Ken Ives of Yorkville Labs. Is the sell-off in tech stocks just a pause, or is it the start of a deeper reversal? We are in touch with Gene Munster, Managing Partner of Deepwater Asset Management. Gene, nice to see you. I know you often say that we are still at the very beginning of the path of AI development.
The market is getting a little nervous about what seems like late- cycle things, at least on the capital market side for AI. How do you interpret this? Well, I'm analyzing what we heard this week , and I think it's important context for this big sell-off. We heard from Meta that they have about 10 million daily active users of Muse, and Zuckerberg is building the entire foundation of the company next year on that.
This suggests that something small, in his opinion, will become very big— billions of users every day in the near future. Of course, we heard news from Tesla and about FSD today. Autonomous driving accounts for about 1% of total ridesharing mileage . The FSD installation rate has increased from about 15% three years ago to 70% today.
Something more significant is happening , albeit at an early stage. We heard from Jensen today. We heard Jensen at GTC in Berlin talking about the power of physical AI , and of course, from SpaceX about Elon's prediction to add 10 gigawatts of computing power next year. By the end of this year, they will have about three.
So my conclusion from all of this is this: if you think these leaders are competent, you have to understand that we are still at the very beginning. Namely, the difference between what happened in the last week and the last three years lies in two phenomena : personalized bots and agents that consume 10-15 times more tokens than a regular chatbot .
And separately—what could happen in the field of physical AI. These two factors together are much larger than anything we have seen in the past few years. So there is a huge gap between what we clearly hear from these companies about the transformational nature and early stage of development, and the market's willingness to accept it.
And I think that in the end, the truth will prevail. And I think the truth is that the world is reorienting itself around AI in ways that we cannot fully comprehend, and that go beyond current high expectations. And these stocks, ultimately, are great to buy now, amid this nervousness, this academic nervousness.
I understand your argument, Gene. You make a very compelling case, but there's a certain transition period until we get to, say, 10 million Muse users, which is a drop in the ocean compared to all the users of Meta platforms, right ? So, I want to say that the potential is there, but it's a very small amount.
And reaching the billion mark is usually the moment when Mark Zuckerberg says you can start monetizing. It's a long way. And at this time, the computational costs are very high. So the question is , is it worth buying stocks in the hope that this transition period will be shorter than we think? That this gap will be bridged by the time they move to the monetization stage.
I'll answer your question with a question : do you believe Zuckerberg is able to predict how things will develop? I think 10 million users is statistically enough data for them to be able to predict how quickly this will start to develop. And despite all the talk about these bots, again, very few people use them.
I use them. I can say that it transforms everything. And finally, the fact that he's willing to come out and say—I forget the exact words, but something like, the center of gravity of this company in a year's time will be Muse. I think this shows his level of confidence. If this were said in a vacuum, if Zuckerberg said this in a vacuum, and we heard other messages from Elon this week or from Jensen, then it would be worth worrying about.
But they all essentially say the same thing . And this "one and the same" is that we have only scratched the surface of the possibilities. I understand that this is a cliché, but this is exactly what is happening now. I don't know when this sell-off will stop, but I do know that the market is expecting a 38% increase in capital spending next year, and the actual figure will probably be 70% or more.
In the 28th calendar year, the market expects 11%. The real figure is likely to be between 25 and 40%. And if that happens, I'm not sure the stock market will react positively to it . I think they'll probably be fine. Gene, I don't think it's necessary to conclude that Mark Zuckerberg and other executives are incompetent, but you can see that all the incentives of capitalism in such conditions push for overbuilding, right?
There is currently a certain shortage. Everyone believes it will be a winner- take-all format or perhaps a duopoly, so everyone wants to be the same, and not one, not two, but three players are competing for it . So I'm just curious to see how this will develop further , because if your projections for capital spending growth next year come true, the market is already resisting such capital needs.
We spend too much time discussing costs. But we don't often think about what the potential really is here. Eventually, Meta started talking less about revenue percentages and more, albeit without numbers, about the importance of commerce and news. And then there will be other productivity apps and entertainment apps .
But that's exactly what we don't know yet. What we do know for sure is that the number of clear examples of AI’s impact on profits is limited. I think Meta is an example where their growth has accelerated significantly over the last couple of years . Google is also an example, but beyond that, there are many experiments going on , but few significant results.
And so I think that's where the gap lies. This, I think, explains why people are so worried about these costs. Regarding your point: yes, the costs will be huge, but do n't they ultimately force each other to spend more and more to stay afloat? The answer is yes, this factor really exists, it is a competitive factor.
But at the heart of this competitive factor lies their own conviction . They don't spend just because their neighbor or competitor is spending, they spend because they believe in it. And that's how it will ultimately be resolved. If they are confident, if they have a good understanding of where the world is heading, then this investment will pay off.
And if they are wrong, then this is the beginning of the end. I believe we are still at the very beginning of the journey. Here's a number that should scare every European leader right now. Europe has only about 5% of the world's AI computing capacity. America has about 75 . 5 vs. 75. In AI racing, it's like coming to a Formula 1 race on a bicycle.
And this week, Nvidia actually said it out loud. Without more computing power, Europe risks being left behind. So in the next few minutes, I'll tell you three stories from Nvidia from this week. One about Europe, one about construction sites, and one about a robot duck. Yes, a duck. And at the end, I'll show you how these three random stories are actually one big plan.
Let's start with Berlin. Jensen Huang took the stage at Nvidia's GTC conference in front of approximately 3,000 people. And his message to Europe was simple: " Hurry up." The problem is in the calculations. This is the " raw" power needed to train and run AI. Europe has talent. She has startups. It has factories.
What it lacks is a sufficient number of data centers. So Jensen demonstrated a system called the Vera Rubin DSK pod. Think of it as an AI data center in a box, built by partners and shipped as a shipping container. Nvidia says this can reduce setup time on a prepared site from months to days. He also said that Nvidia has invested $10 billion in European startups and partners over the past year.
So, Europe urgently needs data centers. But who builds them? And how to build a huge data center without spending millions on mistakes? This is where our second story comes in. There is a company called Procore. It creates software that construction teams use to manage large projects. Procore announced new digital twin features created using Nvidia Omniverse tools.
They are currently in closed beta. So, what is a digital twin? It is a perfect virtual replica of the building before it even exists. Imagine that you first built your house in a video game. You would have noticed the pipe running through the bedroom even before the concrete was poured. Procore Twins link 3D models to schedules, purchase orders, and real-world site photos from DroneDeploy drones.
By the way, Procore is buying DroneDeploy for $845 million . And here's the most interesting thing. Nvidia itself uses Procore to build its own AI factories. But wait, there is a nuance. A digital twin is only as good as the data you put into it . If a contractor's schedule or procurement records are out of date, a duplicate is just a nice invention.
Remember this, because the same rule appears in our next story. The third story is about a working duck. It's called Micro Duck. It stands about 25 cm tall, costs $399, and rolls around on your desk. It is made by Pollen Robotics, which is owned by Hugging Face, and Nvidia has agreed to acquire Hugging Face for about $13 billion.
Reports indicate that the deal is unlikely to close until 2027. So, it's not quite Nvidia's duck yet, but it's on its way to it. Okay, quick game. How many people do you think pre-ordered the $400 robo-duck? Thousand? Five thousand? Record your guess. According to one report, pre- orders have surpassed 20,000 following a viral video featuring Jensen and the Ducks at GTC in Berlin.
20,000 ducks. This is not a toy. This is a whole movement. And she's not just cute. You can teach your duck new movements by first training it in a simulator. For this you will need an Nvidia GPU or the Hugging Face cloud. And now—the big reveal I promised you. These three stories are part of the same plan.
Step one: Nvidia helps design and build AI factories using digital twins, like Procore. Step two: It fills these factories with its chips in places like Europe that desperately need computing power. Step three: It gives AI a body—robots, even as small as Micro Duck, trained in simulation on Nvidia hardware.
And by buying Hugging Face, Nvidia gets a platform where millions of developers share their AI. Creating intelligence, bodies, and community. Nvidia wants to gain share at every level. Do you remember that trick? Digital twins and robots learn from data. Quality data at the input - smart results at the output.
So, let's close the circle. Europe is lagging behind in computing, and Nvidia is pushing it to build faster. Procore digital twins help build data centers with fewer errors if the data is up-to-date and the robot shows the final result. AI that moves in the real world. in one line. Nvidia doesn't just sell chips anymore.
She sells an entire world powered by AI. If you liked it, subscribe. I analyze Nvidia news every week. And next time I'll show you the robot that Nvidia is betting on—and it's much bigger than a duck.
Comments 0