저자: LarryFink · 외국인
요약 보기
래리 핑크는 AI 확산에 따른 전력·에너지 부족과 컴퓨팅 비용 문제를 우려하며, 미국의 에너지 생산 확대(태양광·저장장치 등)를 촉구한다. 오라클은 레버리지·잉여현금흐름 악화와 AI 인프라 투자 부담으로 시장에서 불안이 크지만, 클라우드 인프라 성장과 대규모 수주잔고(약 6000억달러, 절반이 OpenAI)를 근거로 성공 시 큰 성과 가능성을 언급한다. 메모리 반도체는 AI 수요로 강한 가격 결정력을 보였으나 주가가 미래 성공을 상당 부분 선반영했을 수 있으며, 이번 조정이 AI 지출 둔화인지 과열 후 리셋인지 판단이 필요하다고 본다. 메모리 가격의 향후 방향이 핵심 변수라고 강조한다.
다듬은 글
But I am worried about small and medium-sized businesses. How will they be able to compete in this new world of AI technology? So the most important question I'm asking behind the scenes is: how quickly can we reduce the cost of computing? And , you know, they tell me it's like Moore's Law— how quickly they can develop more and more sophisticated analytics to speed up computing.
But until we develop better technologies and systems in our grid to provide more capacity, more stable energy, I believe the United States should be energy neutral. It shouldn't matter to us whether it's solar energy or hydrocarbons. We must be ready for this . But we don't have to...you know, we need to manufacture solar panels in the United States, build energy storage systems.
At least we can make batteries here. We do it. These are my fears. I'm not worried about the bubble. I'm worried that we don't have the ability to build fast enough. And then I watch China. China is building 100 gigawatts of nuclear power. They are building almost 100 gigawatts of solar power. Of course.
They are preparing for this AI revolution and the need for energy. We, we don't do it enough. And it actually scares me when I see states saying they're going to impose a moratorium. This is not an option. The answer is how do we get, how do we quickly give more energy? Why don't we all focus on how to supply more energy, so as not to raise electricity prices, but to become the center of the AI revolution?
No. Can I ask you about leverage? Every morning we come here and check the KOSPI, which has been showing fluctuations of 6-9%. The President of South Korea said that the market there, in her opinion , is quite unstable . A lot of it is credit ETFs and two big names, which we don't really do. Right. Yes.
But does that worry you? Is this a global risk factor? No, I'm not really worried about, uh... AI software development is coming soon, but it's only 21%. I think what everyone is looking at is Oracle's cloud infrastructure, which is about 30% of revenue, is growing rapidly, and that's what everyone is betting on.
Uh, outdated products make up about half, and they're actually pretty easy to protect. Eh, I don't think they're as vulnerable to AI developments as software- as-a- service lines. So it really all depends on Oracle's cloud infrastructure— it's currently 30% of revenue, but it's growing. You know, they have a backlog of orders, I think, about $600 billion.
Uh, half of that comes from Open AI. Everyone is waiting for this IPO, just like Oracle. So all of this, I don't want to use the expression " house of cards," but it's all building up with a lot of stakes for the future, and I think that's what's making investors nervous . Now the question is , at what point does it become too cheap?
Because you know, there's definitely a good chance that they'll succeed, and 118 might seem very cheap in the future. But I think the concern, at least for us, arises when we look at the data on corporate website visits—minus 11%. Their competitors are actually growing. Some of the companies we are very concerned about are showing positive results ; for example, Salesforce.
I mean, we're pretty pessimistic about Salesforce, and yet they're up 8%. So what does this mean for us regarding Oracle with their 11%? And , you know, all these stock charts look pretty tough when you look at them, but being the last in a group of companies that are very vulnerable to future changes is not a good thing.
The only thing that puzzles me a bit again is that only about 20% of what Oracle does is super vulnerable. So I don't think the sell-off is that strong, it's not like Salesforce where I think their entire business model could be turned upside down by software development. But I think it's really just a question of leverage and whether they're borrowing too much?
Is n't their free cash flow being used up too quickly? Capital costs are too high. They are obviously investing heavily in the future. They have bet on everything , and this is where investors have to make decisions right now. Landon, I think this is one of the most interesting stories in the market right now, because the world is allergic to capital spending.
They are allergic to negative free cash flow figures. They need a lightweight asset model and simplicity. And if that were the case, no one would ever build a house. But Oracle is different. They spend money. They take on debt. They are issuing new shares to build physical assets, and that is unpopular right now.
But the bet they're making, Landon, I think, is that this is their element—data. Critical data is what they do best. So, who else is better suited for this? The world hates these numbers now. Hate the fact that they borrow money. Hates negative free cash flow. Hates share issues and capital expenditure figures; this hatred will either disappear or it won't.
But if everything they bet on comes true, it will be a huge success for them . Their chance to completely conquer AI is already on the horizon, if they do it right. Landon, what am I, what am I saying wrong? So I think a lot of investors, especially now, you know, it's a different conversation compared to what it was a year or two ago.
But today, many investors are thinking, “ You know what could go wrong?” And there's a long list for Oracle, right? There are a lot of domino effects that could lead to problems for this company, and as I said, it's a bankroll game for them to some extent . But I think you also have to ask, and if you have a risk appetite, you have to ask yourself: what if everything goes well, huh?
And I think that's exactly the argument you're making. And these stocks, again, have fallen by two-thirds from September to today. This is something important happening beneath the surface of the AI trend. For a month, semiconductor and data center stocks seemed almost unstoppable. They dominated the conversation, memory maker stocks soared, and storage companies soared.
Semiconductor equipment manufacturers have become some of the strongest players in the entire market. Then suddenly the momentum has changed, and when momentum fades from an overheated market, the decline can become as rapid as the rise before it. This is exactly what investors are feeling right now .
The important question is whether this is the beginning of a fundamental recession in AI spending, or simply a major reset after overly optimistic expectations. These are two completely different scenarios, and the evidence now looks much more complex than just a phrase that the AI boom is over. Let's start with memory.
of frastructure. Modern AI systems require vast amounts of DRAM, NAND, and especially high-speed memory as data centers have expanded and demand has begun to outpace global supply, creating extraordinary pricing power for manufacturers. Micron, SK Hynix, Samsung, Western Digital, SanDisk, and Seagate—all of them have benefited from different stages of this infrastructure cycle.
But eventually, stocks began to price in a huge portion of future success. This is where things get dangerous. A great company can still become a failed asset due to mispricing. And the incredible industry could experience a severe correction, even if its growth prospects don't disappear. That's why investors are paying extremely close attention to memory prices right now.
The question is no longer whether memory prices are rising. They have definitely grown. The question is, what happens next? If prices for standard memory begin to stabilize, investors will have to determine whether they will remain high for a long time or whether history will eventually repeat itself.
Historically, memory has been one of the most cyclical areas in technology. Scarcity creates higher prices, and higher prices stimulate investment. Investment creates additional supply , and over time it catches up with demand. Prices fall, make this cycle different. AI has introduced a huge structural source of demand that has never existed on this scale in memory cycles before.
And each new AI cluster requires memory. Each additional GPU requires memory access. Larger models require memory, and longer context windows require even more memory. ependent on memory.
Comments 0