원문 보기
Let's now invite Tom Lee, Fundstrat author and CNBC expert, to the conversation. We find out what he thinks about what he heard. Glad to see you, of course. What do you think about what you just heard from our two guests in the studio? I think both points of view make sense. You know, yields have risen to levels that rival stocks.
I think the market is still trying to understand why yields are higher. I think there are several reasons for this. But two things seem to be lost in this discussion. First, what will the rates be in 6 months, because what will inflation be like? And secondly, I think that increasing profitability improves competition for stronger companies.
This means that some companies will become even more attractive as profitability increases. And I think, as you pointed out, Scott, that's why the Magic Seven is growing. You know, when profitability increases, it's still easy for them to raise funds, but it's harder for competitors. And regarding inflation, I still think the next 6 months will be the most important, because if inflation starts to fall, it won't be because of the Fed's rate hike last week.
This is because of the policies that are already in place, and I think this means that the Fed can tone down its hawkish rhetoric. Yes, but what if that doesn't happen? That yields and inflation will remain high for longer than you expect. The Fed's own forecasts do not promise a return to inflation target next month, or the month after, or later.
Yes, Scott, I think the likelihood of her falling is very high. First, on September 30, they will present a new methodology for calculating PCE. It will likely reduce the rate by 20-40 basis points on an annual basis. So, 3.4% can turn into 3%. In addition, the effects of tariffs and flash memory are starting to fade.
Commissions in the stock market are also being leveled out, and if the price of oil remains at $100, it will not affect inflation in 6 months. So for inflation to accelerate, housing prices would have to rise, but high rates would hold that back, and oil would have to reach $150, and the impact of flash memory would be excluded from the methodology.
So, I think the chances of significantly lower annualized inflation in 6 months are quite high. I know, but there is no sign that we are approaching any kind of balance on the issue of memory. You may have different approaches to the methodology for calculating inflation, but that doesn't mean that inflation itself has necessarily changed.
You simply decided to add other ingredients to the mixture and bake the pie differently. Right. But, you know, for the average person's consumer basket or inflation, the life of the average person does not depend on the prices of flash memory. This, of course, makes AI more expensive, but for the average person, the only real memory cost is the phone in their pocket.
So I really think flash memory prices could still go up. This should not be the reason why the Fed should raise rates. But in a general sense, you're right, Steve. We don't have any indication that inflation is lower, but I would say the likelihood that it will be lower in 6 months is high. And if that happens, the market will finally be able to breathe a sigh of relief.
He has just crossed this significant level. It is now the fourth American chipmaker to achieve this status, after Nvidia, Broadcom and Micron . What does this mean? That's a great question. I think when you look at these four companies, you see exactly where the demand is in the AI market. So Nvidia, obviously, and AMD, which is really being pushed up by two factors.
The first is processors, we can talk about that for a long time , but the other is what AMD is looking at as a second source for AI factories . It's much more than just chips. This is an entire system that companies like HPE and Lenovo are currently building, and they will offer it to customers late this year or early next.
So, I think you're seeing a recognition that AMD is really a second source for companies and advanced AI developers who are building these factories. To what extent was the AMD stock movement we saw this week—I mean, not including today, as a lot of semiconductor stocks are falling— driven by the news, given that Meta is such a big customer?
I think news is a huge driver. I mean, anyone who's used agentive AI, where the AI actually does something for you, not just talks to you, I've been using it at work for almost a year now, and it's become as natural as turning on the lights. What I mean is that it becomes a very important part of your daily life.
And this idea with the introduction of Muse, when people see AI as something that can actually perform tasks, I think it will lead to mass adoption, and that will put a strain on the processors, because they are the conductors of the orchestra . These are chips that tell all the other chips what to do.
So even though GPUs, AI chips, work very, very fast, they need CPUs to control them and give them instructions. Cisco has a great saying about this: they say people click and agents swarm. So you'll see one person with a hundred agents doing all sorts of tasks: going through old emails, contacting old friends.
And I believe that this is what will significantly increase the demand for processors in the future. I want to come back to this in a moment, but first I want to discuss the fact that you're using Muse . I'm curious, you know, I was talking to a previous guest about how this has caused another wave of concerns about the disruptive impact of AI, as we saw yesterday in the financial reports.
Stephen, I'm just curious, in your daily life, when you use this technology, what exactly has it replaced? Well, in the case of Amazon Q, a tool I've been using at work since the beginning, it doesn't necessarily replace anything. It simply makes possible what you couldn't do before. For example, he doesn't book flights for me.
Instead, he plans my two-week trip. And in the era before agent-based AI, I probably wouldn't have studied 15 companies in three cities and all these different possibilities that I could have implemented. And frankly, I would n't have an assistant for this. Their time is better spent on other things.
So I would call it this: it makes possible what previously wasn't worth the time or effort. But suddenly, when it starts happening, it raises the bar for what a good trip, a good presentation, or a good letter to your aunt should look like. Yes, it is definitely a helper in our daily lives. Going back to the success of this technology, what do you think it will mean for the demand for processors in the future?
After all, we have already seen the initial reaction to what this could mean for this subsector. Absolutely true. I mean, Intel says it can only fulfill a certain percentage of its processor orders, because that's what we've all been predicting for a long time: Intel, AMD, and all the other companies that work under the ARM license.
And I think the most important thing now is that software is starting to catch up with the capabilities of the chips themselves. We've seen this in recent evaluations of current- generation chips, where comparisons make it clear that in some cases the software is already fully utilizing the chip's resources.
This is good because companies like Nvidia have switched to releasing a new generation every year, whereas previously a new generation of AI technology would appear every few years. And it all works when we look at members of the “ trillionaire club” like AMD: much of what they’ve announced during valuations in recent weeks suggests they’re using a type of computer-to-computer communication that relies heavily on Broadcom.
So you look at these two stocks, they, they, they're very ...sorry, look at these two companies. You look at these two companies, pay attention to Broadcom, because Broadcom provides a significant part of the communication between these chips. Broadcom also develops a number of specialized chips that use large hyperscalers .
Yes, definitely. And regarding this report—I don't know if AMD confirmed it —that they might have to raise prices. What does this tell you about their pricing power and demand now, as well as about " chipflation"? That's right. Well, I think as a negotiator with many years of experience, I always talk about shortages when I want to raise prices, just to make it clear that it is necessary.
But I think if we look at the increase in the cost of memory—and remember, a lot of these AI factories are running on huge amounts of memory—we've seen a price increase there because of scarcity. So it's only natural that with such a crazy demand for processors, supply can't keep up, and this gives companies like AMD, Intel, and others pricing power that they may not have had with processors a few years ago.
Mhm. You talked about the structure, you know, how it changes. I'm just curious, when you talk about data centers, you say it's a company now, what do you mean by that? I think when you look at a company like AMD, the data center is actually a customer, because remember, these giant systems don't necessarily go to enterprise customers first.
For example, when AMD's Helios comes out, one of the first customers will likely be Oracle, and they are essentially a hyperscaler. So, I think when we look at this, the data center for these large hyperscalers, for Frontier Labs, and then for large enterprise customers—that's really what we're looking at.
It's not so much about selling chips and installing them in servers, it's about creating a holistic, comprehensive solution . When Nvidia's Vera Rubin platform came out, the first ones were shipped to Dell Technologies, I think, a couple of months ago. The cooling system associated with these systems is simply awesome because they need to be cooled.
Therefore, companies need to think about how they will organize cooling around their data centers. So, it really becomes a big infrastructure story, not a product delivery story. There are three stories about Nvidia today, and they all point to the same issue. How strong is this AI boom really as it seems?
The startup is seeking a valuation of $35 billion . Nvidia chips are headed for orbit, and a new report suggests the costs of it all are dangerously concentrated. Let's figure it out. First, Enscale, a UK-based AI cloud infrastructure company, is seeking a $35 billion valuation in pre-IPO funding, more than double its valuation in March.
And Nvidia is directly footing part of the bill, reportedly investing about $2 billion of the raised funds. That's what justifies such a leap. Enscale's contract backlog has grown to approximately $103 billion , nearly doubling in one month , largely due to a 6-year, $45 billion deal to supply Anthropic with computing power.
Nvidia's support of this fundraiser is not charity. Every dollar Enscale spends on data centers translates into GPU orders. So , Nvidia is essentially financing its future customer. Second story. Nvidia and Alphabet are becoming the main beneficiaries of a very early but very real trend— orbital data centers.
Nvidia has already created a chip for this, the Vera Rubin Space Module, designed to provide data center-level AI performance while in Earth orbit. Collaborating with partners such as Star Cloud, Axiom Space, and Planet Labs. The idea is that space offers nearly unlimited solar energy and does not require water cooling , solving two of the biggest limitations to AI development on Earth.
It's still in its early stages and expensive, so skeptics are calling it hype rather than a reality of the near future. But if the scale grows even moderately, this will be another market for Nvidia GPUs that simply didn't exist a few years ago. Third, and this is what connects the other two points, with a healthy dose of caution.
A new report highlights that even as the AI boom attracts new customers, spending remains concentrated in the hands of a small group, directly impacting Nvidia and Coreweave. Coreweave, one of Nvidia's closest partners, has historically derived a huge portion of its revenue from just one or two customers, mostly Microsoft.
This is exactly the same risk that underlies Enscale's new valuation. The $45 billion contract with Anthropic is a huge vote of confidence, but at the same time it is only one step away from the company's huge backlog . More logos are appearing in the AI spending space , but the money is still coming in through a small number of very large checks.
So, the main theme today: Nvidia's presence continues to expand into orbit, into new neoclouds, and higher valuations, but the customer base that funds it all remains extremely small. This is not yet a " red flag" today, but it is this fault line that analysts will be watching if even one of these mega-deals falters. Let me know what you think about it .
Comments 0