저자: JimCramer · 외국인
요약 보기
Cramer는 향후 4일간 시장 방향을 확신할 수 없다며 연준 금리인하 기대가 낮고 SpaceX 딜과 하이퍼스케일러 지분 발행이 자금을 흡수해 단기 부정적이라고 봄. Cerebras 상장 후 급등락 문제를 언급하며 신규 투자자에게 매수 서두르지 말라고 조언. 이후 AI 수요가 공급 제약(GPU·메모리·네트워킹·전력)으로 옮겨가 반도체 강세 논리가 유효하다는 주장이 이어짐. 근거: SOX 지수 상승은 PER 확장보다 실적 성장 주도, 대만 TSMC가 선단 웨이퍼 공급 집중, 'dark GPU' 없음(100% 가동).
다듬은 글
I don't know what's going to happen in the next 4 days. And that can change a lot of things. We now have the snapback from last week. I don't know what could propel us higher still. I do know that a lot of things can go awry, though. So, rate cuts from the Fed are likely off the table. The SpaceX deal will suck money from the rest of the market.
More hyperscaler equity offers could do the same thing. And now Apple's getting its clock cleaned, too? That's more negativity than I can handle. Sure, maybe we're on the gauntlet and go four for four positive. And that could make the market really roar. But the bottom line here is I just don't see it that way.
I think the Fed's more important than anything and it's now going the wrong way. I was hoping for a reversal upward, not downward in Apple. I wanted to get any surprise equity deal over with before the big SpaceX deal. And I fear an insane opening for SpaceX IPO followed by a horrendous decline will leave a real bad taste in everybody's mouth.
So, I am not that bullish. My bullishness can wait. I think you will get a better time to buy than right now. Simple as that. Hey, why don't we go to Angel or Angel in California? Angel. Yes, sir. Good day. Boo-ya, Jim. Thank you, sir, for having me on the show. And I'm glad you called. How can I help?
Well, uh as a genius move, I ended up buying Cerebras uh when it first came out. Mhm. Originally, I thought the stock was coming out of 170, 180. Uh when I went to finalize this close, it was at 370. Right. And it has not been anywhere near that since then. So, I'm wondering if this stock is a dead stock.
Should I get rid of it? Uh today is the first day that it's been Frankly, I think it went up 36 points today. It gave you a nice bounce back. I cut it in half. Then if it goes up any more, I'd get rid of the rest. This was a botched deal. This is what I most fear, okay? They obviously misallocated. They didn't understand that the stock could open very big.
They picked everybody off. And everyone who bought this stock in the last 2 weeks is underwater. And that's precisely, actually three weeks. That's precisely what I don't want to have happen. And I tell you, I've done these kinds of deals, I priced them, I've been in them, and it was done poorly. And now I just heard Angel talk about how he did, and there are hundreds of thousands of people like him and her.
You know, let me just tell you something. I fear that's what's going to happen this week. And it's going to hurt you. And I'm trying to protect you. That's what I'm trying to do, is protect you. Let's go to CJ in Texas. Please, CJ. Hey Jim. Thank you for taking my call. Absolutely. I am a new member of of the club, and I watch your show, and I really thank you for the education we receive.
And I'm a small investor, but I'm a very active investor. We're all small investors, and I thank you for the kind words. Let's go to work. Okay. So, my question relates to ServiceNow, NOW. And I've had it for 5 months investing in what I thought was a growth company that might grow a little faster than it has.
And then the price plummeted. So, I hear mixed projections about the future of holding ServiceNow, and it seems to be mostly a buy, but calls and comments, what happened on Friday? And then everybody saw the Korea trade overnight, which was just a reaction, but they're thinking that's now anticipatory, and that this is all breaking.
What do you tell them? Sure. I I think really what's happening is that the debate has uh fundamentally shifted this year, even more so to the benefit of semiconductors. Uh we had our Bank of America Global Technology Conference last week. Um you know, last year the debate was that if they build it, will they come, right?
This year the debate is, can you build fast enough? So, the debate has completely shifted from the hyperscalers having to justify the return on their investment to uh just constraints everywhere you look. So, whether it's in computing, whether it's in memory, whether it's in networking, whether it's in storage, whether it's in power, and we think semiconductors are a very critical part of helping to address a lot of those uh constraints.
And that's why I think it's not a surprise that the SOX index has done so well primarily driven I should add by earnings growth rather than price to earnings multiple expansion. The reaction I get is from people who go, "Look, did you see that price target of $1,200? That's crazy." I mean what do you say to the when I'm talking about people who've been in the market a long time.
They've seen this game. They've seen this movie before. They know what happens. Everybody chases price targets on the way up. Then there's going to be a move and they're going to chase them down. So and they're they're almost in disbelief that anything meaningful can last longer than the kind of action that we've seen in this strong rally the past couple of weeks.
What what is your message to them? Sure. I I think that to some extent you know the investors who were impacted by some of the historical cycles whether it was a telecom cycle or the build out of the internet. I think they saw a lot of infrastructure being deployed but the utilization of that infrastructure was extremely low.
And right now what we are seeing is that the usage of this infrastructure is exceptionally high. In fact the likes of open AI and Anthropic they are trying to buy building capacity wherever they can find it. You know there is hardly a single GPU out there that is not 100% utilized right. There is no dark GPU.
There is no dark compute. Second I think important thing the industry is doing is planning for multiple years out right. So there is already a road map that goes out for the next two or three years and you have these pools of excellence whether it is in the specific kind of memory chips whether it's in specific kind of wafers.
So for example there is only one Taiwan Semiconductor that is helping provide leading edge wafers to the entire accelerator industry. So when you have that one person who is controlling a lot of this production it is very hard to double order and create that overbuild. So I think what we are seeing is a combination of very strong demand and very disciplined supply coming to the market and that's very different than what we have seen before and I think that's why this cycle is a lot more durable.
Final point I would make is look the three largest companies that I cover Nvidia The next bottleneck could be bigger than the GPUs. For the most of the AI women just focused on one thing, that is GPUs. The conversation was simple. More GPU means more compute and more compute means more AI and more AI means more demand for Nvidia.
That thesis worked extraordinary well, but something interesting is beginning to happen. The bottlenecks are spreading. Today the AI industry is no longer constrained by just GPUs. It's constrained by the memory, it's constrained by the networking, it's constrained by the power. It's constrained by transformer substance and the skilled labor permitting and land and most importantly it's constrained by time.
The industry is discovering that a building AI infrastructure is far more complicated than simply buying ordering the chips. You need the entire stack. This is exactly why Jensen Huang increasingly talk about the AI factories instead of GPUs. A modern AI factory is not a collection of the chips. It's an integrated system networking, memory, cooling, power delivery, software, storage, inferences engine and the security.
And increasingly the automation system management systems. All of these pieces must work together. If one piece is missing then entire system will slow down. That's why some of the biggest opportunity over the next several years may not come from the obvious names. Many investors focus exclusively on Nvidia and Nvidia has absolutely remained one of the most important companies in the world.
But the AI supply chain extend much further. Every bottleneck create a new opportunity. Every shortage create a new winner and we already seen this happen with memory. For a years memory companies struggled with oversupply. Today memory has become one of the most valuable assets in the AI ecosystem.
The The thing is happening with networking. A GPU cluster is the only a effective as the network connecting it. A cluster become larger and networking become more important. Latency matters, bandwidth matters, and scalability and thousand of GPUs are impressive. A hundred thousand of GPUs working together efficiently is transformational.
That's the way networking become critical. Power is the another area investors are beginning to appreciate. AI is becoming an energy story. Every model is required electricity, every inferences required electricity, and every AI factories requires electricity. And the amount of electricity required is increasingly rapidly.
Several years ago, data centers were a niche consumer of power. Today, AI factories are becoming some of the largest energy consumer in the world. That has enormous implication. Utilities, grid operators, nuclear energy, natural gas, power equipment manufacturer, all of them are becoming part of the AI story.
This is why many industry leaders continue emphasizing the infrastructure. The challenge is no longer that AI works. The challenge is building enough infrastructure to support the adoption. That distinction is important. In previous technology cycle, bigger risk was lack of demand. Today, demand is not the problem. Supply is the problem.
Comments 0