원문 보기
On the evening of August 26, Nvidia delivered an earnings report that Wall Street couldn't find a single fault with. The next day, the stock price jumped 8.7%in one go. However, AMD was falling, Micron was falling, Marvell was falling, CoreWeave was falling, and even giants like Meta, Amazon, and Google were all falling.
The world's most profitable AI company saw its stock skyrocket, yet its neighbors all took a collective dive. Why? The latest assessment from Chief Research Officer Tom Lee is that this is not a signal of an AI collapse at all; on the contrary, it’s a major migration of AI capital. Capital is moving from the hardware layer of AI to the software layer, and behind this move, there is an invisible hand.
Mied I mean, I think it's a healthy market, Scott, because they had good numbers, so you want to go up on good news, which it did, which it rarely does. Yeah, and it's rare, so it's breaking a pattern of people thinking people don't care about the earnings. It really did matter, and I think that we are seeing a positive response in software, which of course is on good numbers, and those names are downstream of the AI trade.
So I think it's actually an overall healthy reaction. I know, but the software moves are so specifically related really to what we got from Salesforce and CrowdStrike and Okta and others. I find it interesting that Meta's down, Amazon's down, Alphabet's down, AMD's down, Micron's down, Marvell's down, CoreWeave's down.
I wouldn't have expected that after Nvidia delivered what they did. Yeah, I mean it's also possible people were using those names and not having exposure to Nvidia, so they've got to find a source of funding . CBC hasn't keep what grow louder as you approach election day, now we're still few months away from that.
Yeah, but that feels like it's at play here too, why the market and those specific names, right? Being issue that resonates with voters, and it looks some elections like oh, and then we seeing governors in Republican states and data center states, you know, supporting a pause. So you're absolutely right.
It's become a political issue. I don't know if this is also maybe fed by China which wants the US to slow down AIs. Tom Lee said in the program that Nvidia almost never rises on good news. The script for this company over the past few years has been that it beats expectations every quarter, but every time the earnings report comes out, the stock price actually falls.
It has fallen for four consecutive times after earnings; Wall Street has almost developed a muscle memory that because the earnings are good......it must be time to sell, so this time, many people closed their positions early, waiting to see the results. This time, however, the script didn't play out that way.
Once the report came out, it gapped up, erasing the losses of the past seven trading days entirely. For a stock that is habitually sold off when expectations are met, this is rare. Tom Lee says this shows the market is actually healthy; good news should get a positive reaction, and that is how a normal market should behave.
The host then threw out a soul-searching question: "You say the market is healthy, but why are software stocks rising?""For example, Salesforce skyrocketed 22%, CrowdStrike soared 18% , while Meta, Amazon, Google, Micron, and Marvell are all falling. Since Nvidia pulled off such a beautiful performance, shouldn't the others be basking in its reflected glory?" Tom Lee gave a very elegant and realistic explanation.
He said this isn't selling ; this is capital rebalancing. He stated very bluntly that it is very possible these people have been shorting or underweighting Nvidia all along. How many fund managers over the past two years thought Nvidia was too expensive, preferring to buy AMD, Micron, Marvell, or Coherent, betting on the next Nvidia?
And what was the result? Every time Nvidia released its earnings, it was a slap in their face. With this earnings report, the reality is clear: those Nvidia "alternatives" in your portfolio aren't making money; the only one truly printing money is Nvidia itself, so what do you do? You have to buy Nvidia, but you don't have the cash on hand.
Where does the money come from? It comes from selling other stocks—like AMD, Micron, or Marvell —to free up capital to chase the Nvidia that is actually delivering results. So, it's not that AMD has underlying issues; it’s being drained of capital, effectively becoming an ATM for Nvidia. It’s like holding a bunch of stocks you aren't bullish on, then suddenly finding out stock A is exploding, but you have no position.
You panic and sell off B, C, and D to clear funds to buy A. Why should B, C, and D fall? It's not because their fundamentals have soured, but because they are being used as a source of liquidity. This explains the phenomenon we’re seeing: why the higher Nvidia climbs, the worse other AI stocks perform.
Because capital is a zero-sum game; it’s a rotation, not a rising tide lifting all boats. But Tom Lee's most valuable point isn't that; it’s what he said next: Nvidia’s P/E ratio is only 27.5, and based on forward earnings, it’s just 18. Quarterly revenue is doubling, it’s the most profitable tech giant globally, and it has a forward P/E of 18—do you realize what that means?
A bunch of struggling software companies on the Nasdaq are trading at 30 or 40 times earnings, right? Nvidia’s valuation is significantly lower than the average of the entire AI sector. Crucially, he noted that the P/E keeps contracting; if the stock price stays flat while earnings forecasts keep rising, the P/E drops with each revision, because earnings are currently growing faster than the share price—it gets cheaper as it rises.
They are essentially hinting at a conclusion: as long as earnings expectations keep climbing, the stock price will sooner or later have to catch up, because the valuation has been suppressed far too low. This is why he isn't panicking; he believes this is the core logic of a healthy market. It's not a bubble propped up by sentiment, but a foundation built on real, hard-earned profits.
Now, let’s return to the most peculiar phenomenon: Salesforce is up 22%and C3.ai is up 18% . These aren't just software products; they represent the downstream application layer of AI. Behind this is a major migration of capital, moving from the "shovel sellers"—like hardware, chips, and data centers—to those actually making money with AI, such as software, SaaS, and security applications.
Why? Because the market is starting to realize that raw computing power is useless on its own; it ultimately needs these software and applications to monetize. Nvidia has already proven that AI can burn cash, but can it actually generate profit? Looking at the earnings report from Salesforce, it tells that exact story: AI features are finally beginning to translate into actual revenue.
When the market sees signals that downstream companies can make money, investors think, "Instead of fighting to chase Nvidia's highs in hardware, why not get in early on those downstream players just starting to deliver?" If we add Tom Lee's point, these names are inherently the downstream of the AI trade.
Capital flowing down the value chain is just a normal trend rotation. Think the AI boom is over? Not at all; it’s just shifted to a different battlefield. Now, I want to discuss the second risk Tom Lee mentioned, which I judge to be the most important and easily overlooked point in this entire video.
He says data centers are becoming a political issue, and he even suggests this might be something China wants to see—hoping the U.S. slows its pace. Let's look at what is happening in the U.S. at the end of August 2026: In Ohio, a referendum to completely ban new data centers has passed the review committee and is heading to a vote.
The Florida gubernatorial primary just ended, and the winner is interesting: while taking massive amounts of money from pro-AI crypto and PACs, they simultaneously proposed limiting data center construction in Florida and introduced a taxpayer protection bill, essentially meaning: don't make taxpayers foot the bill for data centers.
On the same day, the Democratic governor of Pennsylvania signed an executive order imposing harsh thresholds on the development of these data centers. The National Republican Senatorial Committee issued a memo warning that data centers have become a "stealth issue" for the entire midterm election cycle—the kind of sudden, unforeseen hot button that catches everyone off guard.
In the polls, the Democratic Party is leading the congressional popular vote by 6.4 percentage points. A recent Pew Research report shows that more than half of Americans are more concerned about AI than they are excited. In 2021 , this figure was only 37%; it has more than doubled in just three years.
There is also a federal-level bill to pause AI data centers, demanding an immediate halt to all new AI data center construction nationwide. From being high-tech factory hubs in New York and California, they have become the " smokestacks" in the neighborhood. They take up land, water, and electricity; as power prices rise, people's utility bills go up.
Public resentment is real and is increasingly erupting ahead of the midterm elections. The US midterm elections are in November 2026. It is now late August, with voting just two months away; this issue is currently in a prime window to be ignited. This explains why asset-heavy, data-center-reliant stocks like AMD, Micron, and Marvell have been performing poorly recently, as every data center they build could now become a target in the election.
The market has begun to discount the expansion of these centers; if new projects are paused, halted, or stifled by local governments, the growth logic for these companies is compromised. Conversely, software companies don't need to build their own data centers; they use others 'computing power and are purely downstream.
They won't be affected by this political risk and may even benefit as AI is pushed toward applications. This is why capital is shifting from hardware to software. Capital isn't fleeing; it’s relocating because it smells political risk. Moomoo is a brand under a US-listed company and section have a chance to get up to $ 100 in stocks and earn up to 8.1%yield on idle cash.
. Now we have two sides: AI valuations are extremely low, earnings are accelerating, the market is rotating healthily, and Tom Lee is also bullish. On the flip side, political risk surrounding data centers is heating up, and the hardware expansion logic is being questioned. After rising on the day, the market fell 4.6%the next day, Friday, giving back $ 250 billion in gains.
So, which one should it be? My judgment is that this is precisely a long-short scenario with maximum information density. For them, both risk factors are very real: one is the politicization of data centers, and the other is that capital reallocation is essentially a zero-sum game. This means that once there is no influx of fresh capital, the market becomes highly fragile; the large bearish candle on Friday is proof of that.
Since it rose 8.7%on Thursday and gave back half of those gains on Friday, what does that tell us? It shows this rally is driven by sentiment-based capital reallocation , not pure fundamentals; once the funding dries up, it will pull back. Therefore, this is not a market where you can just buy blindly; it’s a market that requires a change in mindset.
AI is not dead; it is still here. The logic for profiting from AI has actually become clearer, but it has shifted from a story of broad participation to one where core assets win and laggards lose.
Comments 0