원문 보기
The biggest economic warning for this whole market— whether it's a stock market crash, a global recession, or something else—is not artificial intelligence. And in this video, we're going to use graphics processing units ( GPUs) to understand what's happening in AI. We'll talk about GPUs, AI, and SpaceX, but really we'll look at what's a real " red flag" for this economy and what's really worth paying attention to.
No, this is not a distraction from the financial problems of a small company like Coreweave. This is old news. I'm going to talk about real issues. So let's get to the details. Firstly, this was inspired by a piece produced by TS Lombard, and I actually disagree with a lot of their complaints, and I'll show you exactly why.
And I'll show you how this leads us to where the real problem lies. So, let's focus on that. TS Lombard argues that AI spending is essentially just a reallocation of funds, that we're not seeing much or significant revenue growth yet, and instead we're getting these blue lines that are just capital spending circulating in a circle.
In other words , Google takes some cash flow and spends a hundred billion dollars on chips that go to Nvidia . Nvidia is spending this on some new cloud services as an investment in those services. They buy even more chips from Nvidia, which gives Nvidia more money to invest in another company, and that company invests in an AI startup, creates jobs there, and it all gets recycled in a way.
But where does the fundamental expansion of the economy come from? This is what TS Lombard questions. And to some extent they are right. Capital spending has been a significant driver of gross domestic product in the United States and, to some extent, worldwide. Europe is not holding on so confidently.
But TS Lombard begs the question: where does the revenue come from? And so many people, I think, hold a similar view that this is just some big bubble right now. They argue that the underlying system looks dead, that new cloud services are unprofitable, and companies like Oracle are risky. They further claim that double and triple ordering is likely occurring, and once this normalizes, we will be hit by a " depreciation cliff." And then we're going to have huge problems because we don't even have anywhere to plug in all the GPUs that are available , and then the capital expenditure cycle slows down , and you get an economic problem.
This is a pretty classic narrative from a certain point of view, and I think it's worth starting to debunk some of these myths, and the easiest way to do that is to briefly look at a company like Oracle. We've discussed this before, so we won't go into too much detail, but very briefly: Oracle's net income grew from $2.9 billion to $4.7 billion.
Their net profit is increasing. And while it used to be fair to say that Oracle didn't have enough money to pay its own bills, now their margins are stable, which is a good thing. And I don't own Oracle stock, so I'm not trying to promote them here , but they finally have enough money to pay their bills, which means Oracle is not on the verge of default.
And yes , CoreWeave is a dubious story, but Nebius isn't as bad as people make it out to be. Nebius is a company whose revenue reached 582 billion here. The only reason they show operating losses is because of the write-off of the cost of chips. $259 million . So, if we add back depreciation, Nebius is actually profitable.
I also have no financial interest in Nebius. This is not such a big problem. The argument is simply that Nebius and Oracle are not on the verge of default, and AI is not going to suddenly collapse because of them. That's not the problem at all . So I think TS Lombard is somewhat bringing them here to say: that's it.
It's all going to fall apart because of these two. I think there's actually a much better indicator of what's happen Nebius and Oracle. It's completely related to what's happening with 6.5- year-old chips and the depreciation gap, because Michael Burry, along with many others, including myself, have been saying regularly that chips will get cheaper at some point in the future .
Just as airplanes should lose value, chips should also lose value. But the problem is that with the recent advent, well, maybe not the advent , but the recent explosion in popularity of open-source models , we're seeing old chips that are 6.5 years old, 4.5 years old, and 2.5 years old getting more expensive, not cheaper.
And this actually means that if Oracle and Nebus, which they pointed to, are not flying into the abyss, then that's not the problem. If the depreciation curve is not actually falling, but rather rising, then this is not a problem. Then DS Lombard is looking for the cause of the economic collapse in the wrong place .
And that's the whole point. We are going to figure out where the main danger really lies. But first, let me prove why talking about a crash due to depreciation is too premature. And I understand what a depreciation curve is. Ultimately, this will be part of the bubble bursting—the story of depreciation.
But it hasn't arrived yet. Why? Look at this. Now there's a way to invest in the future advertising revenue of YouTube bloggers on their channels. This is quite impressive, as we know that social networks are developing at a breakneck pace, and YouTube channels are a real business. They bring in real money.
People get paid monthly, and historically it has been difficult to find authors to invest in. It's interesting, and they are the sponsors of this segment. Go to meetkevin.com/igastar to learn more, but here's the gist of how it works. Gigastar allows creators to sell a portion of the future revenue from their YouTube channel.
These shares are called Channel Revenue Tokens or CRTs. These are securities, not coins or cryptocurrency. Gigastar is registered with the SEC and FINRA, they even created a platform for trading Gigastar. So, if you are buying a share of a YouTuber's future potential income and want to sell it to gain liquidity in 12 months, you can potentially use their secondary platform .
The company itself is also raising funds on Wefunder . This is wefunder.com/gigastar. This is an investment in Gigastar, not in the channel. If you go to meetkevin.com/gigastar, you can learn more about what this company does. This is a startup, so of course there are risks in every investment you make, whether it's in a YouTuber or a company, and you should always read the investment documents.
So check it out at meetkevin.com/igastar. This segment was sponsored by Gigastar. They pay my company Reinvest $4,000 to create it. Check it out at meetkevin.com/igastar. The story with depreciation has not yet come. Why? Look at this. This is a chip that is 6 and a half years old, it is called the Nvidia A100.
Rental prices for the Nvidia A100 chip were decreasing until the end of 2025. But in 2026, something changed. What happened is that we got co- working from Claude and generative AI, which became practically useful, and the changes in productivity are just starting now. You see, someone said this to me, uh, I was in an interview with Yahoo Finance and I said, "I'm more worried about the job market collapsing because the improvements that we're seeing in agent-based AI are just now starting to destroy jobs." Not in 23, not in 24, when the AI was still hallucinating.
Er , not in 2025, when the AI was running into memory limitations. AI wasn't capable of taking over the job at that time. AI has just become capable of taking over jobs. And so someone said to me at the end of this Yahoo Finance interview, "Well, I think if AI was going to take jobs, it would have done it by now." And that's why I want to write this down.
If AI were going to destroy jobs, it would have done so already. This is some kind of financier, like a "financial bro" from Wall Street. If a Wall Street financial bro tells you, "Oh, well, don't worry about the job market." If it was meant to happen, it would have happened already. "then you should be very worried, because I think they are deeply mistaken and blind to the fact that agent AI is becoming so powerful that you won't see this economy collapse at the GPU rental rates at Neoclouds and SpaceX." SpaceX's superintelligence revenue has literally tripled thanks to the rental of their chips.
Good. Much of this comes from leasing equipment purchased for their capital expenditures, but their down payment deposits have increased by 35% or more, meaning that SpaceX's pursuit of hyperscalers is not just Elon's exaggeration. This is actually reflected in their financial statements. They are getting checks for billions of dollars to continue to ramp up cause the problem is not that hyperscalers don't consider AI productive.
Listen, the 6.5-year chip shows rental rates rising in 2026. This is unusual. This 2020 chip , sorry, should depreciate to zero. It should be heading towards zero. It does not go to zero. It is actually getting more expensive, which is an anomaly. The reason for the growth is that these open-weight models allow more agent-based computation to be performed on these older chips .
Therefore, it makes sense to rent these old chips when there are still not enough chipsets to run all the agent-based AI that is currently in operation. That's why OpenAI and Anthropic are demanding any chips they can get their hands on . And this is really a 2026 phenomenon , which is too early to see the impact on the labor market.
In fact, I would say that agent-based AI only became really good, maybe, aside from programming, around July. I really think this is where agent-based AI got, uh, really good. And if we take the period from December to now, we see an increase in the cost of renting GPUs for these chips from approximately 135 to 152.
This is an increase of 12.6%. Growth of 12.6%. This is the opposite of the depreciation curve. You might say, " Okay, Kevin, but this is a 6 and a half year old chip." What about a chip that is 4 years old? Like the H100, which everyone knows about, it was released in September 2022 . Here, please. This is the H100.
It became significantly cheaper as new technologies emerged. This is your normal curve. This is the "I agree with Michael Bure, we're done for " curve. Okay, the pain of depreciation is coming. Oh, but wait a minute. What the hell is happening again in 2026? Oh my god, it's growing again. We are now up 15% compared to the fourth quarter of 2025.
What is happening? Why are H100s getting more expensive again? Productivity is only now starting to be realized. The AI boom due to productivity is just beginning. And this creates a real first signal of danger. But friends, these are not just four-year-old chips. These are not just 6.5-year-old chips.
What about chips that are 2.5 years old? Oh my God, damn it, Michael Bure. You are wrong again. Because, frankly, being wrong about time is the same as being wrong about everything. And that's exactly what we have here . This is simply wrong. Now you are wrong because the rent for Frontier Blackwell's GB200 chips, not even GB300 or Vera Rubin, has increased by 39% since the end of 2025.
This is madness. Everything college football in the chat , you are absolutely right. This is another indicator that artificial intelligence is not slowing down at all in terms of development. Take a look at ASML, which is your very first leading indicator of what's happening. Their order book is exploding.
It's surprising to see an almost threefold increase in orders for cars worth half a billion dollars. These are machines that manufacture wafers from which chips are created. ASML's order portfolio is growing rapidly. This is not a sign that artificial intelligence is winding down. That's why I say TS Lombard is wrong.
It's not AI that's killing this market. I'll tell you what to watch out for, and even though it's still early, I want you to know about it because it's better to know what to look out for sooner rather than later. So, before that, I'll spend quite a bit of time on a little advertising. Go to reinvestone.com if you agree with my point of view.
Join us, because this is the last year you can get lifetime access to all of these products. terminal, real estate terminal, research terminal—all of this is coming later this year. Nine courses, live broadcasts for course participants, agent modules for consumers and businesses. All of this is coming out at the end of this year, except for these AI modules, which will come out in early 2027.
You get lifetime access if you buy this and you will never have to pay a dime on reinvestone.com again. Starting next year, we will introduce an annual fee for participating in this. So you better hurry in 2026, and our next discount coupon expires on October 30th. So be sure to join us. It ends in about 9 days.
If you have any questions, you can always email us . The email address is on the website. Just go to reinvestone.com. OK. So , what is the real problem we need to address in the economy? It sounds a little wild, but it's the unemployment rate, specifically what's happening to the unemployment rate among the black population.
The New York Times and USA Today have made the argument that, typically , when unemployment among black workers rises, it could be an early signal of trouble for everyone. And the black unemployment rate just showed the fastest increase we've seen since 2012. In other words, we saw the fastest growth rate in a single month since 2020, not counting the COVID period, of course, when unemployment rates soared for everyone.
So, they argue, we are starting to see some employment difficulties for black workers, who are often the last to be hired but the first to be fired. In fact, Jerome Powell told us in 2028, "Hooray, we're finally seeing the expansion of the workforce into other, less privileged segments of the economy." He was essentially saying, " Oh, we...don't call me." "Sorry, this is so annoying." Some random number on my phone is calling me.
And I'm usually always on "do not disturb " mode. I can't put it on "do not disturb", but then the computer starts ringing too, which is incredibly annoying. We need to turn this nonsense off immediately. I'm on "do not disturb " mode. God, don't they know I'm trying to record a video? How rude. Apparently, they are looking for work.
No. OK. So, back to the topic. Jerome Powell was very happy that black employment rates were rising at the end of the decade, and then, of course, COVID came. And now there are fears that the unemployment rate among blacks is rising again. And this is happening simultaneously with a number of other events.
Not only the unemployment rate among black people, but also Netflix, which is cutting 6 % of its staff. This was news just a few days ago. is cutting 9% of its staff. This is news only this morning. Of course, layoffs always happen. There will still be millions of layoffs every year. I mean, you have a labor market of 153 million workers.
So there will always be layoffs. What worries me is that at some point our ability to absorb these layoffs will disappear. And I think, this is just my opinion, given what we're seeing with artificial intelligence, AI productivity is about to skyrocket, which is bad for a lot of workers. Actually, don't just take my word for it.
Look at this note right here. I do n't even know how to pronounce this company's name, but this research firm, Nat Natis or whatever, claims that the September labor market report was disappointing, with jobs growing by only 29%. They argue that the job market is actually more volatile than the headlines suggest, and that things are generally not bad right now.
But overall, this rate hike cycle from the Federal Reserve is likely to have fewer hikes due to the state of the labor market. Remember Kevin Warsh's comments about the job market? Kevin Warsh told us that we need to pay attention to tail risks. This is the risk of the left "tail" of the downturn. This would be the tail risk that unemployment will be higher than expected.
He specifically mentioned the labor market and the fact that there is no inflation on the labor market side , which we also saw in the retail sales data, in the latest labor market report and in the ADP report. The ADP private sector report showed us that professional and business services are struggling in the labor market.
Where is the growth coming from now? From construction. From the construction of data centers. But it's not all economics. If it weren't for the construction of data centers, the construction industry would be significantly negative for the unemployment market or the labor market. So we have to be very careful with this.
And that makes me raise my bets that we'll probably only see one rate hike this year. I think the Fed is well aware that this labor market is not showing wage inflation. The Fed cannot raise rates to stop the development of AI. The Fed cannot raise rates to stop the oil industry from growing. All they can do is raise the rate to control inflation expectations, and they already control them.
They are stable after the last increase. And I believe the Federal Reserve is unlikely to continue raising rates in an uncertain labor market. This doesn't mean that the labor market will definitely collapse, but I personally believe, given what we're seeing with agent-based AI, that the labor market has very big problems.
I'll quickly give you a few examples. I'll show you a video about it. I just want to point out that I'm betting on this, and I tweeted about it yesterday . Follow me on X if you haven't already , but I have a few hundred dollars. This is not much. This is more for fun, but I bet a few hundred dollars on just one raise because I think the odds will double to 50% after the Iran deal.
But I also believe we can get softer labor market data, so even without a deal with Iran, if the data is weaker in early November than in October or in early December, I don't think the Fed will raise rates again in December. So I think it will grow. That's just my bet on it. I like to see what's on people's minds, and I don't think many people really pay attention to the job market, so I like to point that out .
Now, obviously, we already know that. Khi is a sponsor of my startup. So if you go to kshi.com/r/mevin you can get $25 when you sign up for Koshi and make your first trade. You know, don't take my word for it about what to trade. ee, you don't have to copy me, but again, they are a paid sponsor and you can use my code Kevin when you sign up to get that 25 bucks.
But I want to say that I think these betting markets, these prediction markets, are suggesting that people are underestimating how much weak pricing policy is hurting the labor market. So when we see a weak retail sales report or a weak labor market report that says, "Oh, there's no labor market wage inflation," that means that when the labor market price action is falling, people are grabbing any job because they don't have the luxury of choice.
The New York Times and USA Today have written about how so many people are now using AI to write job applications or for hiring that many companies are simply giving up, saying, "We don't even care about these applications . We'd rather hire someone we know, if we need anyone at all ." In my opinion, based on what we see in our startup, many people's jobs are under greater threat than the market is estimating.
Mainly because I'm starting to see the real possibility of artificial intelligence taking jobs, or I should say, jobs away from accountants. You can literally entrust Claude with your accounting right now. Of course, subject to verification by a certified auditor or someone else. You can entrust AI to do graphic design for you, work on a website, or edit videos.
This is just madness . Let me see if I have it ...Yes, it's here on my desktop. I'll run this in the background. I had about 30 minutes of footage, and that's one of the reasons I don't usually post many family videos, because editing them takes a lot of time, and after filming my family, I'm the last person to want to sit around editing them.
So I usually only keep the full versions. But I said, "Hey, cut this 30-minute video down to a three-minute clip," essentially turning to the AI. And all the assembly glue that you see, almost all of it, I would say about 95%, is the work . So, AI transcriptions, AI editing, and I think normally if I sat down to do it myself, it would take me an hour to do the whole editing.
Instead, it took about 3 minutes, because you just upload the video, make a request, come back in 20 minutes, watch it, and go, "Oh, not bad." A little tweak here, tweak there, export is ready. This is scary. This is a job for which, you know, you no longer need to hire a video editor. This is madness.
And this is just a quick example that is n't even work-related. That is, it is for the family. But this, you know, is pretty incredible. By the way, we were walking around and watching , you know, Halloween demos. People were putting up Halloween decorations or something , and it's pretty cool. But in any case, this glitch is only in the editing—this didn't exist three months ago.
This happened just in the last 3 months. That's why I say that the damage to the labor market is not yet here, it's ahead. This is a real "canary in the mine" that needs to be monitored. It's not about selling everything because the economy is about to collapse. No, no, no, no, no, no. Artificial intelligence will continue to evolve.
And that's why I'm very optimistic right now, because I actually think we're going to see this huge...well, we have all this nervousness around the Iran deal or something like that. I actually think we're going to see a huge euphoric rally in the stock market after the Iran deal. Everyone's going to say, " Oh my God, look how much money they're making on AI and this and that," right?
And then, at some point in the future, real data will emerge that will show that the labor market is starting to give up ground. Some early signs are already emerging, and I think things will eventually get much worse. So that's my biggest concern for the future. And that, I hope, explains why I'm really optimistic now, at least for the next year .
I think we will make good money in the stock market over the next year. After that, we will start to see some of these signs manifest. Labor force participation rates are normalizing. The bets are off , man. The labor market, because that's how it all happens . Once the job market starts to decline, what will happen?
The labor market will decline. Interest rates will start to fall. But not only will rates start to fall, the stock market usually falls too, because people start to get nervous: "[ __ ], if the unemployment rate goes up, we could go into a recession." Let's withdraw money from stocks. That's when capital spending funding slows down, because it's no longer possible to sustain that massive drain on funds to give Elon everything he needs to implement his plans.
Now you can do it. So the labor market is fine now. Once it goes into decline, the funding will disappear. That's when the capital spending cycle will start to slow down. So the real indicator of problems is the labor market, because it is what triggers the slowdown in capital spending , not the other way around.
There are no signs of a slowdown in capital spending . On the contrary, there are signs of acceleration in the field of artificial intelligence. That's why, by the way, people ask, "Kevin, why, why do these agent AI modules exist?" "Well, because we've found really amazing ways to use artificial intelligence, especially AI agents, and we don't think many people are doing that or know how to use them." So I wrote here, in this little note to the TS Lombard piece, that they are asking the question: are these revenues coming from businesses and end users from outside the system, probably to pay for productivity tools, or simply through the recycling of capital expenditure .
We believe this is mostly a recycling of capital expenditures . Maybe that was the case, but I think it translates into productivity tools that are bad for the job market to begin with. In the long run, the labor market will create new jobs, just like any industrial revolution before it, but initially it will lead to layoffs.
But anyway, I wrote, “We have found huge and incredible ways to improve productivity with AI.” I think most people just haven't realized this potential yet. That's why part of the Reinvest terminal launch will include agent-based AI for business. That is, tips on how to grow your business with the help of agentive AI, and, of course, the consumer module of agentive AI.
They will be available in early 2027. And remember , you can get lifetime access at reinvestone.com. So I hope you'll join us, because it's one payment and you get everythi.
Comments 0