저자: JimCramer · 외국인
요약 보기
엔비디아 중심 AI 반도체 랠리 점검. Dell 급등과 소프트웨어주 반등 속 엔비디아·마이크론 조정 원인 분석. 마이크론은 HBM 수요로 메모리 사이클이 길어질 수 있다고 주장. CapEx 대비 수익 괴리와 투자 인내심 약화를 경계하지만 AI 인프라 수요는 여전히 강하다고 봄. 엔비디아는 6월 Computex 발표 기대감 제시.
다듬은 글
for the hardware computer space with one of the biggest blowouts I can ever recall. On a day where the Dow jumped 363 points, the S&P advanced 0.22%, the Nasdaq gained 0.21%. Dell with a 102-point gain or 32% one session defined today, getting still one more day of tech strength even if we did see some recent winners get hit with a bizarre round of profit-taking at the end of the day.
SELL SELL SELL SELL SELL SELL. DELL'S QUARTER was a true surprise and Michael Dell should be very proud. But there's one great tech quarter after another in May that we saw great results. And nearly all of them were connected to the same thing, the data center in some way, shape, or form. Today, away from Dell, we actually saw some stunning strength in a host of software companies that have underperformed of late.
I'm talking about stocks like ServiceNow, Atlassian, Salesforce, Microsoft, all which were simple they gave you some magnificent performances as if nothing were ever wrong with them and they'd been punished almost a whole month. Their gains were in some ways as shocking as the decline of some of the great hardware leaders, but in the last few minutes of today's session.
At Dell, they saw incredible strength across all their businesses, but the fastest-growing segment came from Dell's partnership with Nvidia, the AI pioneer with a stock that's been a bizarre underperformer of late, including today where the company's stock fell four points in the last few minutes of the trading session and finished down an ugly $3.
It was a wild and inexplicable fall for the largest stock in the world. I didn't think it'd collapse so fast on no news. No news at all. Maybe that's a good place to start our game plan then because the week begins on Sunday night when Jensen Huang, the CEO, gives his annual Computex talk in Taiwan.
Maybe something he says could reverse today's inexplicable decline. This year, Jensen plays host with some of his favorite partners including Renee Haas and Arm Holdings. That stock held up. Matt Murphy from Marvell, Intel's Lip Bu Tan. That stock did not hold up. And Qualcomm's Cristiano Amon. Computex has been a stake in the ground really for many years now for Jensen Huang.
And I think he may unveil something special here. Now, why do I feel that way? We got a potential hint when Nvidia's official X account posted a cryptic tweet teasing quote a new era of PC. The account for Microsoft Windows posted the same message. Grasps at straws doing everything I can for you. Maybe what we saw in late day trading was some sort of end of the month rebalancing out of some semis and into depleted
The S&P 500? Yeah, you know, there's not much sort of interest rate sensitivity specifically for where the trajectory of earnings has been concentrated in that broader AI space. Arguably, you do potentially run into problems as you go down into the smaller companies that have less earnings or maybe don't have earnings at all and struggle to pay interest payments.
So, I think we'll see some differentiating at the cap level to some degree depending on what rates do. But the most important thing for the equity market is not simply just a move by the Fed toward tighter policy. It's the speed that they have to tighten policy that matters. If the Fed, once they move to a tighter stance, can take a slow escalator up and move slowly with rate hikes versus if they have to take a fast elevator up.
The the former is a much better backdrop for equities. What has happened in the past though in the lead-in to the first rate hike, cyclicals tend to do well. We're seeing that now. But once the Fed does start hiking, you tend to see a little bit more of a focus on defensives. So, that's just another way to think about market dynamics as we approach the potential for a tighter Fed.
And when you mention AI, I want to have a conversation about CapEx cuz obviously there's a huge driver right now. I mean, as I'm looking at Amazon, Microsoft, Alphabet, and Meta, each growing their CapEx by about 70% year-over-year in 2025, as you look at that, Lizanne, are we seeing the markets basically giving these management teams a blank check for AI infrastructure or are you starting to see now that investor patience is is beginning to wear thin?
I think there are starting to be concerns that creep in and I think that helps to explain the what I've been calling the great chip dip recently. And it's around a a confluence of things, some of which is shown here. You know, the amount of capital that is consumed from a capital spending standpoint is significantly higher than the profit contribution as you see the share of net income and this is relative to the S&P 500.
So, that gap is something to be mindful of. You also have increasing number of companies, just corporate enterprises saying that they may have to start to consider pairing back on some of the spend because we're not yet fully seeing the concrete nature of either productivity benefits or in the case of capital spending giants, the return on invested capital.
So, I don't think we're in anything resembling lookout below, but those concerns are starting to creep in and I think there's a more discerning eye on a lot of these data points. And what's And what's your thoughts on that narrative that we start to see some of this CapEx spending slow down if we don't soon see monetization backing it?
I I think you will start to hear maybe more than just anecdotal stories of kind of weighing the pros and cons of of a continued massive AI spend even with you know, the price of tokens coming down, the spend is still significant. One interesting way to think about it though, and I've had conversation with corporate conversations with corporate leaders, is they're starting to say at what point do we look at the cost of human capital as now being relatively inexpensive?
So, if if we think about the concerns about the disruption to the labor force associated with AI, there may be some benefit there. Today, we need to talk about what's happening with the two of the most important companies in the AI trade right now, Nvidia and the Micron. Because something ver. Micron recently delivered a extremely strong earning report.
Demand for memory remained robust, and companies continue spending billions of dollars on AI infrastructure. And managements continue to talk about the strong visibility into the future demand. Yet, Micron's stock has fallen more than 20% from its recent high. At the same time, Nvidia continue to trade below its previous high, despite the fact that the world still appear to be heavily constrained when it comes to AI computing capacity.
So, the questions investors are asking is simple. Is the AI trade is beginning to weaken? Are we simply watching a normal correction after the one of the most strongest rallies semiconductor industry has ever experienced? Let's start with Micron. Because I think the situation surrounding these companies explain exactly what investors are struggling with right now.
For the decades, memory business has been known as the one of the most cyclic part of the semiconductor industry. The cycle was relatively simple. Demand would increase, memory prices would rise, and companies would make enormous profit. And those profits would encourage manufacturers to build more production capacity.
Eventually, that additional supply would enter the market, and price would fall. Margin would collapse, and the investors who bought memories company near the top of the cycle would suffer enormous losses. That history is exactly why Wall Street has traditionally refused to give companies like Micron high valuation multiples.
The market simply didn't trust the earnings. Whenever Micron generated enormous profit, investors assumed those profit would eventually disappear. But today, the argument surrounding Micron is beginning to change, and the reason is artificial intelligence. Memory is no longer simply being used to store footage, photos, and the videos applications, and traditional consumer data.
inside the modern infrastructure AI infrastructure. And a 1 GPU without enough memory cannot operate at its full potential. The more powerful AI models become, the more data they process. The larger the context window become, the more inferences workload increase, the more memory the entire ecosystem requires.
That's why high-bandwidth memories has become so important, and that's also why the current memory shortage would potentially last much longer than the previous cycle. The semiconductor industries cannot simply increase production overnight. Building a modern semiconductor fabrication facility requires billions of dollars.
Then companies need to construct advanced clean rooms, install extremely complicated manufacturing equipments, hire engineers, bring production, and the qualify those product with customers, and finally ramp the manufacturing. That entire process can take years, and high-bandwidth memories create another problem.
Producing HBM requires significantly more manufacturing capacity than producing traditional memory.
Comments 0