저자: TomLee · 외국인
요약 보기
결론: 톰 리는 AI 성장 기대와 금리 완화 전망을 바탕으로 주식 시장(S&P500 8,200~8,400 가능성)과 기술주 상승을 낙관하며, AI 규제 논쟁에서는 혁신 억제를 우려하고 있다. [거시·금리] 최근 고용보고서가 이전 8월 수치가 예외적이었음을 시사하며, 시장이 연준의 3차례 인상(10월 인상 확률 75%)을 반영하던 매파 심리가 완화됐다고 주장. 향후 6개월간 인플레이션이 둔화되고 연준이 매파 입장에서 벗어날 것으로 봄. 10년물 금리는 잭슨홀 이전 4.75% 수준이며 5% 아래면 위험자산에 긍정적이라고 언급하고, 6개월 내 5% 아래 가능성을 제시. 금리 정상화는 인플레이션 불확실성 감소와 연결. [주식·지수] S&P500이 연말까지 약 9% 상승해 8,200~8,400 수준 가능성. 시장은 '압축된 스프링'이며 AI 주도 상승으로 지지받는다고 평가. 3분기 실적은 약 30% 성장 전망, 2027년 EPS 상향 조정이 연초 대비 20% 이상으로 S&P 상승률(20% 미만)보다 커서 밸류에이션(PER)은 오히려 낮아졌다고 주장. [기술주·암호화폐] 기술주가 신고가를 경신하고 암호화폐도 강세, 긴축 환경에서는 어려우므로 금융여건 완화를 예상하는 신호로 해석. [AI 산업·규제] AI 기업 주식의 거품은 영원하지 않으며 가격 하락과 일부 기업 파산이 예상되나, 닷컴버블 이후처럼 기술 혁명으로 이어진다고 봄. 샘 올트먼, 다리오 아모데이 등 AI 기업 경영진이 공공정책을 주도하는 것은 이해충돌이라 비판하고, 규제는 기존 강자의 해자를 만든다고 주장. 다만 범죄·해킹 등 실질 문제는 법으로 규제가 필요하다는 입장도 함께 언급. 기업 간 과도한 보조금·토큰 경쟁 축소와 오픈소스 모델 확대 가능성도 언급. [스페이스X·데이터센터 관련] 스페이스X 분기 매출 약 146억 달러(전년 대비 161% 증가)가 기대를 상회, AI 컴퓨팅 사업 확대를 발표하며 2027년까지 약 10GW 컴퓨팅 능력 증설 목표 제시(고성장 시나리오 20GW). 이는 엔비디아 등 반도체·인프라 수요에 긍정적이나, 전력·냉각·네트워크 등 병목으로 GPU 대기(다크 GPU) 문제 발생 가능. 수요 부족이 아니라 인프라 지연이면 배치 지연 리스크는 있으나 치명적이지는 않다고 평가. [리스크] 금리 재상승 또는 인플레이션 재가속, AI 과열 조정, 전력 인프라 지연, 규제 강화가 주요 변수. [출연자 전망] 톰 리는 기술주·S&P500 강세, 금리 하락 가능성, AI 인프라 수요 지속에 대해 낙관적 입장 유지.
다듬은 글
Uh, that unemployment or employment report that you think means the previous one in August might have been a bit exceptional. And from this you conclude that any inflationary pressure from the labor market is unlikely. So, again, you think that if oil is brought under control, inflation might cease to be a problem.
Yes, good morning, Joe. That's right. I think that looking at last week's employment report, the market was operating with the most hawkish sentiment on the Fed going to raise rates three times, including a 75% chance of an October hike. This employment report was softer. Um, we'll get the inflation report in 2 weeks for September.
In our opinion, yes. We think these one-off changes are happening, so I think there will be softer inflation readings over the next 6 months, and that will allow not only the Fed to move away from its hawkish stance, but also allow bond yields to actually normalize as inflation uncertainty starts to diminish.
And I think there are a lot of clues, because I think tech stocks are hitting new highs and cryptocurrencies are doing so well, but they can't do well in a tightening monetary environment. So, I think they expect an easing of financial conditions ahead. So you look at different markets, what's happening.
That would be important, Tom. So when you said yields are normalizing, what do you think the normal yield on the 10-year bond would be right now if there weren't these, what you think, fakes in different markets? I think one number we can work with is that before Jackson Hole, the 10- year bond was 4.75%, and I think anything below 5% would be interpreted by the markets as really positive for risk.
And I mean, there's a good chance that in the next 6 months we'll see yields below 5%. You also expect the S &P index to rise, perhaps as much as 9%, by the end of the year. So you have maybe 8200, maybe 8400, and you call the current market, which I think, again, you, you go against conventional thinking all the time.
You think this market is a compressed spring, ready to rise much higher. It's almost as if he's being held back, although most people think he's getting by because his skis are only being lifted by artificial intelligence. Uh, yeah, I think we're living in an unprecedented time because earnings growth is accelerating, which is hard for a lot of people to believe in their lifetime, but even third-quarter earnings are probably up almost 30%.
And the P/E, sorry, EPS, revisions from the beginning of this year to now for 2027 are up more than 20%, and the S&P is not even up 20%. So the multiplier actually went down. If anyone thinks the market was expensive at the beginning of the year, it has actually become cheaper. Uh, and then I think ironically, oil prices are rising even after some of the biggest AI companies have pledged to rein in their own growth, but AI stocks have also helped the markets reach record highs.
Can self-regulation really work when everyone has an incentive to cross the finish line first? Well, I think it might. And first, let's be frank. The froth you see in the stocks of artificial intelligence companies won't last forever. Hmm. There will definitely be commercialization of this technology as more and more people compete for the best technology, the best type of artificial intelligence, the best bot, the best Muse.
Muse is the latest artificial intelligence invention to come out of Facebook. So , you will get commercialization. You will have to expect prices to fall. These promotions will not be as successful. Some artificial intelligence companies are going bankrupt. This is exactly what happened during the dot-com bubble.
But think about what happened next. Google, Microsoft, you know very well , Microsoft has already been there, I would say. Google, Facebook, I can go on and on. You know, we had Amazon. We had a technological revolution that improved our lives. It will happen here. And this is where it gets interesting, and I think this is where the rubber meets these, you know, AI geeks who are trying to set public policy.
They, Sam Altman and Dario Amodei, am I right? Amodeus. Amodeus. I have to...I have to get the Italian name right. Um, they shouldn't set public policy. They have their own axes that need to be mixed up . They have a conflict of interest. They have ...They are the first. This means they have a head start over everyone.
They should try. They compete. Regarding Taylor's topic, I mean, these companies are competing. And Taylor, you said who would be the first to finish? They don't want to compete. You see, this is what they want to stop. They want to stop innovation. And here is my final remark. Regulation. Well, because they like regulation...
Big Big Banks like regulation... You don't think Jamie Dimon likes regulation, do you? To say that regulation creates a moat so that existing players can win and all other companies can win. Let's be frank. If we don't advance this technology, we won't be able to fix its glitches. Okay, here they are , the glitches.
There are problems. I'll just quickly...I'll just quickly say... I mean, the market will reward the company that produces the most reliable, the most affordable, the safest...That's right. So, when you say about the police themselves, it says that they, like you, are governed by the law. The government is still enforcing contracts.
He still defends individual rights. But let companies innovate, not innovators, not bureaucrats. Yes, but, you know, there are still serious crimes. Yeah, and all that, you know. But we already see these companies doing these things. There's no doubt that AI will be the most transformative technology of our time, but we're already seeing it, you know, I would say, committing crimes.
Now we say that when an Anthropic agent hacks a company, it says that an Anthropic agent hacks XYZ. But you can throw out the agent. It's Anthropic hacking this company. I use OpenAI every day. If I used ChatGPT to hack the Australian government, I would be behind bars. We don't see it now because it's in the testing phase and they're putting it off, saying they're operating outside the realm of possibility.
But we have laws regarding these types of regulations. We do n't see this yet. And lastly, when it comes to these companies, you need to think about what they want. They want to make a lot of money, and they...And how does that not happen? Competition and whether they can be sued. So regulation prevents things like open-weight models from being a bigger force in the industry.
Now, yes, the best models and prices will drop. But the models are heavily subsidized. Previously, we saw a tokenization race where companies encouraged their employees to use as much as possible. Tokens are expensive, so when we stop subsidizing these, um, these different...You can get different subscription levels.
OpenAI or open AI is becoming much more attractive. So I think they want to ...They're afraid of open- source models, and I think they're also afraid of lawsuits and regulatory protections... Before we get started, on the cost of data centers and whether the AI boom can continue to grow at the same pace.
But SpaceX's new data raises another question. What if the AI industry doesn't meet demand? What if it runs out of power. At the same time, reports that GPUs are waiting for power have sparked heated debate about Nvidia's future. So let's see what's really going on. What the latest numbers could .
SpaceX's performance, which caught investors' attention, reported SpaceX revenue for the quarter of about $14.6 billion, up 161% from a year earlier. This exceeded analysts' expectations by about $12.7 billion . Its space most interesting business has been its artificial intelligence and computing business, which the company now calls its harp increase from the previous that supplieat is directly attracting Nvidia's attention.
. And this is another great impressive comment came when Elon was discussing the company's plans for the future. He said SpaceX aims to have a much larger ambition viewed as a management forecast rather than add sustainable computing power with the goal of increasing capacity by approximately 10 gigawatts by 2027 .
Musk also discussed a high-ambition scenario that includes 20 gigawatts of additional power and cooling capacity, while acknowledging the difficulty of achieving it. Why is this important to Nvidia? Because every significant expansion of AI computing requires more than just software. It will need processors, memory, networking, electricity, cooling, and the infrastructure that connects everything together.
If SpaceX follows through on its plans, during the SpaceX earnings call isn'. Consider . It can't just install 1,000 , and the network equipment must be installed. The memory and servers must work together, and the entire system must be tested and put into operation. If one key component is missing, some computing power may have to wait.
suggested any facility had enough available power to run them. The report describes GPUs that should have been in use but could not yet be enabled. Some people call them dark GPUs. At first, it can't use . Equipment waiting for electricity is not the same as equipment waiting for customers . If a company has no demand for its computing power, unused GPUs can become a serious financial problem.
But if customers are waiting for computing power and the company is still building the energy infrastructure needed to provide it, the situation is completely different. The first scenario indicates weak demand. The second indicates a delay in the launch of deliveries. This does not mean that delay is harmless.
Equipment that is not being used can create problems, and companies must carefully manage the time of their investments.
Comments 0