저자: LarryFink·JamieDimon·MorganStanley관계자등다수출연 · 외국인
요약 보기
결론: 영상은 미국 경제와 AI 투자 사이클을 낙관하면서도 재정적자·금리·지정학 리스크를 경고하는 내용이며, 대상 자산에 대한 구체적 1~3개월 가격 방향 예측은 거의 없다. [거시 경제 진단] 미국 경제는 현재 상승 국면이며 기업 이익이 크게 늘었다고 언급됐다. 기업과 소비자 모두 과도한 부채 부담이 없고, 실업률은 매우 낮다고 평가됐다. 저소득층 임금 상승은 충분하지 않을 수 있다는 단서가 붙었다. [자본시장 및 재정] 자본시장으로 자금이 몰리는 추세가 강조됐다. 미국 국가부채는 2000년 약 8조 달러에서 현재 약 40조 달러로 26년간 약 32조 달러 늘었다고 언급됐다. 국채 발행 규모가 커지면서 자금 수요가 늘고 있어 대출 가능 자금 수요에 영향을 준다는 취지다. 또한 민간 자본을 유치해 미국 경제를 연 3% 성장시켜야 한다는 주장이 나왔다. [3분기 흐름과 4분기 전망] 3분기 미국 주식시장은 보합, 채권은 3% 하락이었다는 언급이 있었다. 7~8월은 활발했으나 9월에는 인플레이션, 금리 상승, 유가 상승으로 분위기가 바뀌었다. 출연자는 지정학 상황이 다소 안정되면 AI 효과, 성장, 인플레이션 통제, 양호한 기업 이익이라는 긍정 시나리오가 복원될 수 있다며 4분기에 조심스럽게 낙관적이라고 밝혔다. [AI 슈퍼사이클] Morgan Stanley 측은 AI 투자 사이클이 세대에 한 번 있을 현상이라고 봤으며, 현재 사이클은 10~15% 진행된 것으로 평가했다. 하이퍼스케일러 자본지출이 연 1조 달러 이상이 될 수 있고 장기적으로 10~15조 달러 규모가 거론됐다. 데이터센터 정치적 반대, 금리 상승, 보안 우려에 따른 투자 둔화에도 불구하고 기대치는 유지된다고 밝혔다. Morgan Stanley는 이 과정에서 자문·구조화 역할을 맡아 시장 점유율을 두 배로 늘렸다고 했다. [SpaceX 및 AI 인프라] SpaceX 분기 매출은 약 146억 달러로 전년 대비 161% 증가했으며 시장 예상치 127억 달러를 상회했다. AI 컴퓨팅 부문 매출은 약 75.6억 달러, 조정 EBITDA는 약 88억 달러로 언급됐다. Musk의 연간 매출 목표는 12월 약 1000억 달러, 2027년 말 5000억~1조 달러이며 이는 경영진 전망이다. 2027년까지 약 10기가와트 추가 컴퓨팅 용량 계획이 제시됐다. [병목과 GPU 유휴 논란] AI 병목은 칩에서 전력·냉각·건물 등 인프라로 옮겨가고 있다는 분석이다. Amazon 보고서에서 일부 Nvidia 시스템이 전력 준비 전에 인도돼 '다크 GPU' 논란이 나왔다. 출연자는 이것이 수요 부족이 아니라 인프라 지연 문제일 수 있으며, 지연 비용·감가상각·금융비용 리스크는 있으나 곧바로 AI 버블의 증거로 볼 수는 없다고 봤다. 핵심은 인프라 완공 후 장비가 충분한 수익을 내는지 여부다. [종목·자산별 발언 정리] Nvidia: SpaceX가 향후 컴퓨팅 인프라에 Nvidia 플랫폼 사용을 약속했다는 언급이 있어 수요 측면에서 긍정적으로 소개됐다. 다만 가격 방향에 대한 명시적 예측은 없다. 삼성전자, SK하이닉스, 금, SOXL, TQQQ, 비트코인, 브로드컴, S&P500, 테슬라, 팔란티어, 마이크론, 나스닥, 코스피, IEF, 미국채금리, 한국채금리, 환율, AMD 등 나머지 대상 자산: 영상에서 1~3개월 가격 방향 예측이 확인되지 않았다. [리스크 요약] 미국 재정적자와 국채 증가, 금리·인플레이션 재상승, 지정학 리스크, AI 인프라 지연, 데이터센터 정치적 반대가 주요 하방 요인으로 제시됐다. [출연자 전망 요약] Fink·Dimon·Morgan Stanley 측 모두 장기적으로 AI 사이클과 미국 자본시장 성장을 낙관하지만, 단기적으로는 지정학과 금리 변수에 따라 분위기가 달라질 수 있다는 입장이다.
다듬은 글
Look, the American economy is on a bit of an upswing right now. These are good times, you know, corporate profits have increased significantly. And the good news is that I don't like to make assumptions about the economy. I like to analyze the factors that drive it, and then you can assess the probabilities.
Corporations do not have excessive debt burdens. Consumers also do not have excessive debt. Unemployment is very low. It is debatable whether wages are rising enough for the poorest, probably not, but they have jobs. And that's the single most important thing for most people— to have a job and the opportunity to work .
And most people, contrary to what you hear, like their jobs. There is a steady demand for capital in all segments of the economy. You know, every quarter when I'm here, I talk about the impressive nature of the U.S. capital markets. And we see more and more of the economy going through the US capital markets .
I mean, even yesterday we announced a big deal with Humana. Uh, you know, in the field of elderly care . Last week we announced a very big deal with AT&T. I know that everything related to artificial intelligence and data centers is exciting, and we want to talk about it . But more and more organizations are coming to the capital markets.
BlackRock is at the center of these conversations. You know, I like to say about our quarter and year: markets drive our assets, but our customers drive our growth. And, you know, our growth this quarter: if you remember, the US stock market was flat in the third quarter, bonds were down 3%, and we were up 205 billion.
So, we see more and more companies approaching us for funding, and that's the trend we're seeing. But let me take a step back, because I keep saying this. Yes, we are now at the stage where we are starting to open up private equity. Jim, you and I are constantly shouting about this , shouting about opening up private markets, opening up this sector.
But at the same time, the US budget deficit, which we don't talk about enough, is starting to have a serious impact. You know, we've come a long way from $8 trillion in national debt in 2000, accumulated over 223 years. And national debt, I mean national debt, not deficit. Yes. The national debt is now 40 trillion.
We have grown by $32 trillion in 26 years. So, let's talk about this as a huge impact on the demand for loanable funds. Our government needs more and more loans. Bond auctions are becoming larger and larger. At the same time, I already said, I think, last quarter, that we need to attract private capital to grow the US economy by 3%.
You have, in essence, the largest pool of capital in the world. I have also noticed that more and more international investors are coming to invest in the US. But, Larry, there is a concern that what is driving a lot of our economy and our markets right now is that this capital spending... that it's not enough.
Right. Capital expenditure rates, which we have never seen before in our careers, remain low. Two-thirds of our economy is moving ... How about this idea that there might be a capital shortage? That some firms, maybe smaller than yours but still very large, are saying, "You know what?" "I've already exhausted my data center funding limits." And this can become a limitation.
That's why I'm saying that we have to open up new ways to finance this. You know, as kids, Walter and I started in 1976 . I was doing something called mortgage debt, which no one had ever heard of. We started it, and back then it was, you know , a few billion dollars. Now it is 15 trillion. Two phases of the quarter.
There was a lively July and August, as you said, and September changed the equation because of inflation, higher rates and oil. Do you expect the fourth quarter to be more like the beginning of the third quarter or the end of it? Well, that's a great question, isn't it? Two of the known unknowns this year were the phenomenon of AI and geopolitics.
And it seems like the lights just went out, but we're continuing, right? But they came back. Here we are. Okay, the glass is half full. The interaction of the phenomenon of AI and geopolitics has been a dominant reality throughout the year. And what we saw last quarter, actually in two quarters, as you say, the first quarter was risk-oriented.
The AI phenomenon stimulates GDP growth by 5-6%. The economy continues to grow at full employment. The Fed is focused on inflation expectations. The mood is positive. And then, as Labor Day approached, people started asking about the pace at which capital spending would be done. Are interest rates and inflation higher, more persistent, and such that they actually indicate some additional risk to the economy?
Well, and of course, geopolitics plays a role. So these are actually two very different stories. And I would like to think that we will have the first scenario, a positive one, if we can at least take some control of geopolitics . If the geopolitical situation stabilizes somewhat, I think a lot of things will return to the positive aspect, namely—that the AI phenomenon is real.
The economy continues to grow because of this. Inflation is under control. Corporate profits are excellent. So I'm cautiously optimistic about the fourth quarter. Where I am more optimistic is regarding Morgan Stanley's role in this process. Because we act as the primary advisor to clients, both in wealth management and investment banking, and in markets where we are very busy.
I want to ask about the AI supercycle because, as you mentioned, it has become critically important both in terms of sentiment and for Morgan Stanley's business. During last quarter's report, you said that we were only 10-15 % through this cycle. Since then, we have seen greater political opposition to data centers.
We saw higher interest rates, as you mentioned. And we see big players slowing down their own development due to security concerns. Have your expectations changed given these obstacles? I mean the longevity of this supercycle and the ability to profit from it. You are asking the right question. We must accept the assumption that the AI supercycle is a phenomenon that occurs once every generation, or even once every few generations.
And we at Morgan Stanley believe that this is so. So the question is, at what rate will hyperscaler capital expenditures roll out? Is that a trillion a year? Is it one and two trillion a year? Will it take eight or ten years? Will it be 12 trillion or 15 trillion? These are huge numbers, but over time, I think this is the reality: if you want to be in the game and win, you have to invest large amounts.
And a significant portion of these flows are directed to the companies with the best capital and the largest cash flows in the world. And our role is to act as an intermediary, helping to structure private and public funds, debts, and shares. Leslie, I think, without a doubt, the AI investment cycle is here .
How long does it take to implement, will investments sometimes outpace technology, or vice versa? Certainly. This is the case with any phenomenon that happens once in a generation or even less often. But for efficiency, productivity, and sustainability, the digital sphere is already here. And I think it's critical that Morgan Stanley be part of this as a trusted advisor.
Therefore, we doubled our share in this direction. And it doesn't matter whether hyperscaler capital expenditures are a trillion next year, a trillion two hundred billion or 900 billion, what matters is where we're headed. Where we're headed is probably in the long term, depending on how you measure it, between 10 and 15 trillion, which is a big number compared to the S&P of 7 or 8 trillion.
But it's worth remembering that this provides, perhaps, 50 basis points of real growth in the US economy, as you know. So, it's real, and we believe it will last. the new AI in the stock market . Something unusual is happening in the AI market . Investors are concerned about this. They question the costs of data centers, and some wonder if companies have already built too much AI infrastructure.
But at the same time, big tech companies continue to compete for more computing power. SpaceX reported explosive revenue growth. Amazon says demand for cloud capacity remains difficult to meet. In addition, the next generation of AI systems is becoming closer to the consumer. So, what story is the market missing?
Are we nearing the end of the AI boom, or are we witnessing the early stages of a much larger computing industry? Let's start with SpaceX, as it's a fresh result that offers an interesting look at where the AI market might be headed. SpaceX is becoming a bigger player in the AI space. SpaceX reported quarterly revenue of approximately $14.6 billion, up 161% from the same period a year ago.
This was significantly higher than the market's expected figure of about $12.7 billion . Its satellite internet business, Starlink , has shown strong results. Its space business has also exceeded expectations. But the segment that has attracted the most attention is its AI computing business, now called the superintelligence segment.
This business generated about $7.56 billion in revenue for the quarter. According to the report, the company also reported adjusted EBITDA of about $8.8 billion . This figure shows why investors are paying more attention to SpaceX's computing plans. It doesn't just launch rockets and build satellites.
It also works to provide the computing resources needed to run advanced AI systems. And we are right in the middle of this opportunity. SpaceX has committed to using Nvidia's platform for its future computing infrastructure. If the company expands as planned, it could become an increasingly important source of demand for AI chips.
But the most ambitious part of this story is what happens next. During a revenue discussion, Elon Musk said his SpaceX aims to reach about $100 billion in annual revenue by December. He also outlined a potential path to annual revenue of between $500 billion and $1 trillion by the end of 2027. And this is the real goal of the owner.
This is also a management forecast, not a result that has already been achieved. Its achievement will depend on execution, customer demand, infrastructure expansion, and the company's ability to convert planned capacity into a profitable business. Still, the scale of ambition suggests something important.
Some of the biggest players in technology believe that the demand for computing for AI could grow significantly in the future. The next " bottleneck" is electricity. This is where the story gets interesting. For a long time, conversations about AI focused mostly on chips. Will AI be able to produce enough GPUs?
Will manufacturers be able to supply enough components? Will advanced chip factories keep up with orders? Now the industry faces another challenge— energy. A company can purchase thousands of advanced GPUs, but they won't bring in revenue until the entire data center is up and running. This means that electricity , cooling, networks, buildings, and other equipment must be in one place.
SpaceX management emphasized these challenges when discussing its expansion plan. Musk said the company plans to gain about 10 gigawatts of additional computing power by 2027. He also discussed a more ambitious scenario involving 20 gigawatts of additional power and cooling. The largest goal, according to his comments, will be difficult to achieve, but even a 10 gigawatt plan would represent a significant expansion.
Why should AI investors care? Because every new data center needs equipment, and if companies are delayed due to a lack of energy rather than a lack of customers, demand for chips could remain high even if installation takes longer than expected. And that's a completely different problem than having too much equipment and not enough people willing to use it.
Are unused Nvidia GPUs a red flag? The Amazon report raised concerns that some Nvidia systems were delivered before certain facilities had enough electricity to operate them. In simple terms: some of the equipment was ready before the infrastructure around it. This has led to discussions about so-called "dark" GPUs, meaning those that have n't even been enabled yet.
At first glance, this sounds like a serious problem. After all, why would a company spend billions on computing equipment that can't be used immediately? The answer depends on what exactly is causing the delay. Imagine a company building a huge data center. Servers arrive in October, and electrical equipment becomes available in November.
The company's system is being completed in December. Servers may be idle for some time, even if the company already has customers waiting for computing power. And this creates a problem with timing. But this does not necessarily indicate a demand problem. Of course, delays still cost money. Equipment may depreciate, financing costs continue, and revenue may arrive later than planned.
Investors should not ignore these risks. But they also shouldn't consider every unused GPU as evidence of an AI "bubble ." The critical question is whether the equipment will be operational and generate sufficient revenue once the supporting infrastructure is ready. If so, the delay may be temporary. If not, investors will have to consider a much more serious problem.
Comments 0