원문 보기
Now, one of the largest companies on the planet has lost more than 12% from its record high , and the bubble warning index has just skyrocketed. If you own a broad U.S. stock index, you own a share of Nvidia, which means it's about your money. I looked through everything that was posted this week and it made me stop and think.
Because the same week that the headlines were screaming about a correction, the Nvidia CEO said something that, if true, would make this panic look much less significant. Who would want to diversify their investments when AI data center stocks are so hot? And only AI data center stocks. Nothing more.
Everything else is icy. But when this group of stocks collapsed over the past few weeks, only diversification and cash could save your portfolio from destruction. If you focus too much on one thing, you will fail. You might abandon the entire asset class and move into cash , as many did in 2000, 2008, 2022.
Those who left left trillions on the table. My goal is to make sure this doesn't happen to you. We're wrapping up today's CNBC Investing Club meeting, which was delayed because we had to discuss the "slow-motion disaster" that the tech sector has become . Almost every day for the past month, we have seen major sell-offs in data center-related assets.
It was a real bloodbath, led by SK Hynix, the Korean memory chip maker that dominates the Korean market, and it's starting to have the same impact here. After soaring nearly 1,000 % in the two years ending in June, SK Hynix, along with its memory-maker cousin Samsung, have become a disaster. How bad is everything?
SK Hynix, the one traded in the US, was at $195 10 days ago . It is now worth 146 bucks. Light. In recent years, the Korean market has been gripped by a shopping frenzy unlike anywhere else . Private investors dominate the Korean market, unlike us, and they don't seem to care at all about diversification.
I've seen statistics showing that SK Hynix and Samsung make up a huge portion of traders' portfolios in Korea, where many own only those two stocks and nothing else. This is real overcrowding. These stocks have simply been mowed down over the past few weeks. I mean, they're " mowing down" the shareholders.
Especially those fools who bought them with borrowed money. Unfortunately, there are some consequences here, although this only worsens an already bad situation. These two Korean stocks, like most names in the tech equipment space, desperately need the AI data center story to stay relevant. Perhaps, but there are some cracks in this story.
Building data centers requires huge amounts of money, but rising interest rates have made it harder to get the money . This is quite logical. The two largest companies in the field, OpenAI and Anthropic, still remain private, although the latter could go public by Thanksgiving. Last night, Anthropic released sales figures and a forecast that some even called disappointing.
Meanwhile, CoreWeave, a heavily indebted data center developer we often talk about, has faced an attack from short sellers who claim the company may have trouble continuing to build because its public debt has a yield of 15%. Well , that's quite worrying, even though its private land loans are trading much higher.
However, CoreWeave's weakness is nothing new. This has been going on since the stock peaked in May at 138. It's now around 69. This is the worst week of the year. But it was the collapse of its bonds, not its shares, that made people panic, as CoreWeave has huge debt, much of it secured by Nvidia semiconductors.
So, let's understand this chain. For a long time, the entire data center ecosystem has been growing rapidly. Koreans were increasingly actively buying shares on our market. They raised prices for the most profitable semiconductors, as well as, by the way, Tesla and Nvidia. When things were going well , oh, everything was good.
But now that the collapse has occurred, the Korean sell-off has directly affected our market. Concentration and lack of diversification caused a massive sell-off in semiconductors. I said the situation has gotten so bad that you can get a sense of how badly our tech stocks will be hit just by getting up early and watching SK Hynix and Samsung trade overnight.
At the end, you'll learn why Nvidia fell, whether it's a warning or a gift, and how to read this news without losing your head. And I'll show you one number this week that weighs more than every red candle on the chart. It received much less attention than the fall itself. First, about the worst. Today's Benzinga headline: "Nvidia enters correction after rising wedge breakout as AI stock bubble index jumps." A rising wedge is a pattern where the price continues to rise, but the ups weaken, like a compressed spring.
Traders take this as a bearish signal , although it is never a guarantee. This time he broke through to the bottom. In a few weeks, Nvidia fell from a record high of $243 to around $214. A correction simply means a drop of at least 10% from the peak. This bubble index has risen to 74 , which is a high number, as the extreme zone starts at 80.
This is not a crystal ball. This is an assessment based on six indicators, including how expensive AI stocks are and the gap between AI spending and revenue generated from it. But when a warning light comes on just as your stock is falling, you feel anxious. I felt it too. But wait, the graphs only show people's sentiments.
They don't talk about what exactly the company sells. Remember this. This will become important in about 3 minutes. Look wider. Earlier this year, Apple joined Nvidia as one of only two companies in history to reach a valuation of $5 trillion. So, a $ 5 trillion company is in a correction phase, and the bubble index is flashing.
House of cards, right? Here's the thing: stocks move based on earnings, expectations, and external forces. This fall has a little bit of everything . Chipmaker stocks were sold off across the market. Some analysts point to rising bond yields. Coreweave, which rents out computing power based on Nvidia chips, fell more than 15% in one day.
And there are reports that some AI companies' revenues have been lower than expected. These fears are real, but there is another side to the coin. I read Jensen Huang's comments on this week's podcast and laughed because he sounded almost defiant. To paraphrase, he said Nvidia is growing its share because it offers more for global capital spending .
Capital expenditures are funds spent on building something. And the world's largest companies are spending staggering sums on AI data centers. His statement? Nvidia is capturing a larger share of this than before. And now the game. Data centers are measured in gigawatts, just like power plants. For every gigawatt built, Nvidia has the ability to sell chips, networking equipment, and systems.
Guess how much it costs. Record the number and we'll see how close you are. First, a valid fear. Specialized chips, called ASICs, created by Google, Amazon and others, are gaining market share. So why should you keep paying Nvidia? Here is the twist. Nvidia has long ceased to be just a GPU manufacturing company.
A modern AI data center also needs CPUs, networks that make thousands of machines work as one brain, and ready-made AI systems. It's less like selling a single ingredient and more like selling an entire kitchen. Maybe it's just a CEO promoting his stock. Rightly. So, here are two places where the idea of " kitchen" manifests itself.
Nvidia RTX Spark laptops are available from six manufacturers starting at $2,400, featuring a Blackwell graphics chip and Grace processor on a single superchip, and Mercedes-Benz and Nvidia just announced an expanded partnership in driver assistance with an eye toward future level four autonomous driving, though neither company gave a timeline.
Data centers, laptops , cars—is this expansion or collapse? But expansion is not guaranteed. A valuation of $5 trillion leaves little room for disappointment. That's the real risk, not the wedge. Here is your answer. Nvidia says its potential revenue per gigawatt has more than doubled, to over $40 billion.
Careful, this is Nvidia's own estimate of what it can sell, not guaranteed sales or profits. Jensen estimated it at about 18 billion for Hopper, about 25 for Grace Blackwell, and over 40 for Vera Rubin. That's more than a doubling in two generations of chips. If you guessed a lower number, don't worry.
This is easy to miss while everyone is looking at the graphs. Is Nvidia a good buy? I ca n't say. I am not a financial advisor. Bubbles exist, and stocks can fall even when businesses are doing well. A smart person takes both versions into account , asks what is already included in the price, and decides according to their own plans.
So, let's summarize. The fall is real, and it's more than just a mood. The bubble index is a warning signal , not a sentence. The figure was 40 billion per gigawatt—this is Nvidia's own estimate. If you remember one sentence: a falling price shows how nervous people are, but what Nvidia plans to sell shows the goals of the business.
And it only matters when it shows up on income statements. If this was helpful, please subscribe. And then there's another company in the $ 5 trillion club, Apple, and at least one analyst says it looks more expensive than Nvidia. I want to know why.
Comments 0