저자: ScottGalloway—주출연자SteveEisman · 외국인
요약 보기
AI 기업(OpenAI·Anthropic) 적자와 빅테크의 AI 매출 집중도가 큰 리스크라는 주장. 빅테크 잉여현금흐름 감소, 감가상각 증가 언급. 중국 저가 모델에 의한 가격전쟁이 촉발 요인이 될 수 있으나 시점은 불확실하며 현재 신용·고용 지표는 양호. 시장은 현재 강세이나 'AI 가격전쟁' 신호 전까지는 상승 지속 전망. 호르무즈·관세 관련 물류 논의도 포함.
다듬은 글
If something goes wrong with OpenAI and Anthropic, which are losing billions of dollars, it will all fall apart. Losing billions is putting it mildly. If only they were losing billions. Billions, yes. That's right . They lose much more. So, I mean, the question for me is , how likely is this "if"? And at least I would expect that the probability of that " if" is priced in to some extent , but when I see the S&P trading at record highs...
Yeah, but that's not how markets work. The market will get a blow to the head because it has been " bullish" for too long and everyone is just buying on every dip. In other words, it is now purely reactive, rather than proactive in predicting what might happen. He will not be proactive at all . You know, Michael is trying to predict.
God willing . Welcome to Frothy Markets. I'm Ed Elson. Today is August 6. Let's take a look at yesterday's market performance. The major indices were mixed after a series of record highs, more on that in a moment . Brent crude was relatively stable, Treasury yields were unchanged, and finally, Google shares fell nearly 4% on news that DeepMind's chief scientist was leaving and CEO Demis Hassabis was stepping down.
OK. What is happening? The S&P 500 just hit another record high. Yesterday's new intraday high followed Tuesday's results, when the index rose nearly 2% to its first record close since June. The Nasdaq also added nearly 3% that day, and the Dow crossed the 54,000 mark for the first time. It seems that investors in the markets have shed the worries about AI that defined much of the past month.
But some people don't believe it, and that's Michael Bury. In a note on Tuesday, the investor who predicted the 2008 crash said he stands by his bets against Nvidia, Micron, Tesla, Palantir and the semiconductor index. He said, quote, "I continue to believe that we may be close to a serious peak," adding that we could see, quote, "a 1987-type slump." Now investors are asking themselves who is right—the bears or the bulls.
To discuss this, we speak with Steve Eisman, legendary investor from The Great Short Sale and host of The Real Eisman Playbook. Steve, thank you for joining us on Frothy Markets. Glad to be back. You're one of those guys who predicted the crash of 2008, one of those guys from that movie . Yes, it will be written on our tombstones.
There is nothing to argue about here. That's right. I mean, what do you think about this market right now? Because I thought we were all worried about AI, about debt, about dependence on a few AI labs, but we're at record highs . So what is the market really telling us right now? I can give you the bearish arguments or we can discuss it, but right now , first of all, the US economy is very strong.
You know, all banks reported in mid- July, and credit indicators were as favorable as possible. So overall, there are no credit problems in the US economy. It's a strong M &A cycle, the IPO calendar is good , and certainly the economy is still K- shaped, but if you look at Visa and MasterCard data, overall payment volumes are quite strong.
Where you really see a K- shaped economy is, for example, in Procter & Gamble, which has revenue growth. You know, that's not what drives the economy. Right now, everything is, well, okay. No, despite all the hysteria about AI destroying every job on the planet, employment figures are still very, very strong.
I think that's why the market continues to grow, because everything is just normal. Now I think the story of AI has become much more complicated. If you and I were sitting here a year ago, it would have been a complete "hooray!" That is, it would be difficult to find anyone who would say anything negative.
Maybe someone like Gary Marcus, who was on my show, but he was probably on yours too. I adore Gary. Gary has a lot to say, and that's all great, but last July, Gary Marcus was about the only person on the planet who said anything negative about AI . Although there was also Citron. What I mean is that the problem is compared to 2008 .
Yes. So, Michael and I had the same thesis: mortgage underwriting standards had deteriorated dramatically. And what was good about this thesis was that every month, securitization data came out , demonstrating the credit quality of hundreds of billions of mortgage loans. So you could see, you could really see the situation getting worse every month.
You had an incredibly robust data set that came out every month and confirmed, "You 're right." So you were getting that kind of reinforcement. You know, if you're going to make up a negative story about AI, you don't have a securitization database to help you . So, what, what negative story do you have?
Let me tell you what I consider to be a potentially negative story, but it hasn't happened yet. Yes. So, this is what makes the AI story more complicated. First, there is no doubt that this business is much more capital-intensive than anyone could have imagined. Companies like Microsoft, Google, and Amazon, which once generated cash like water, now have negative cash flow.
Yes. This is a big change. This does not mean that these companies are on the verge of something bad. It just means that the dynamics of their business have really changed. They are investing huge amounts of money. Whether they will make a significant profit from this, we don't know yet. But this is one big change.
Another major change is that the business of agent-based AI models, such as Anthropic, OpenAI, and several others, has no “ moats” because people are constantly switching from one model to another. And now open models from China have appeared, which are much cheaper, and people seem to be switching to them.
Maybe there will be a price war, maybe not. Where I think there's a vulnerability: I read a report that said about 70% of Amazon and Microsoft's capital spending is on OpenAI and Anthropic. Yes. This is exactly what I'm looking for. If a massive price war starts over Chinese models and OpenAI and Anthropic get into big trouble, it will destroy a significant portion of the AI market because they won't be able to spend as much, and therefore will buy fewer chips.
I think you can figure out the rest yourself. But until that happens, these companies continue to waste money like water. So, Nvidia will report, it seems, on August 26th. This should be a good quarter. I couldn't... How could they have a bad quarter when Amazon is spending $225 billion this year on capital expenditures ?
So, until we see real weakness, I think, from OpenAI and Anthropic. I'm not on Barry's side. Well, it's very interesting because, as you mentioned , we're seeing more numbers on the reliance on OpenAI and Anthropic, at least among big tech companies. Just today, Bloomberg reported in its own analysis that 70% of Microsoft's AI revenue comes from OpenAI—a company in which Microsoft has, of course, invested.
So, I think it's fair to say that Microsoft is investing money in OpenAI, and that money is coming back to them in the form of AI revenue, which is literally the majority of their entire AI business. Yes. You raised an important issue. You know, if we see a price war, if something goes wrong for OpenAI and Anthropic, who are losing billions of dollars, then this whole thing will fall apart.
" Lost billions" is a euphemism. If only they were losing billions. Billions, yes. That's right . They lose much more. So, I mean, the question for me is, how likely is it if...Uh, it's quite likely for me at this point. It seems to me that everything is going in this direction. And at least I would expect that the probability of this is already factored into the prices to some extent .
But when I see the S&P trading at record levels... the market is working. Markets do n't work that way. I mean... You know, the news is still ...the economy is still in good shape. If there is a price war, the market will have to be hit on the head with a heavy object, figuratively speaking. Because it's been a "bull" market for so long that everyone just buys on every dip.
In other words, this is now purely reactive behavior, not proactive prediction of what might happen. This will not be proactive at all. You know, Michael tries to predict events. God help us. He has more courage than I do at this point, because I just think it's premature for me . I am waiting; if it happens...and I mean, you think it's very likely and I wouldn't necessarily disagree, but it could happen in a year.
That's the thing . I mean, if it happens in two months—that's one thing. If a year passes, all these companies will be spending money the same way they spent it before, and it's the same story. So, assuming that Anthropic and OpenAI do get into trouble, the big question, the real question, is when? How long will it take?
And I don't think anyone...I definitely don't have the answer to that question, and I don't think anyone else has the answer right now. True. How do you think other investors on Wall Street, how much do you think the market is asking this question? I agree with you that no one has the answer. I don't have an answer.
You don't have an answer. But you and I, at least, are asking this question, which instills a certain indecision or at least anxiety in the entire ecosystem. Do you think people ask this question, or are the numbers so exciting that no one cares? You know, I would like to answer this question. It would be nice to get all the investors in a room, have group therapy, and then we could get an answer.
Good survey. But otherwise, you know, I do n't know. I...I absolutely just don't know. I mean, you know, I'm even a little surprised by the fact that I thought that when the 10-year bond went above 4.5%, that might have been the limit. And that didn't happen. You know, if the market is going to get a correction because of rates, I think the 10- year has to go above five.
So, this is a "bull" market. You know, all of this is being discarded. I mean, I'm surprised because I thought his numbers last week were ridiculous. Simply terrible. And the stocks, you know, fell for one day, and then went up again. What part of their performance worried you the most ? Costs? Two...two groups.
So, the first is a 28% increase in revenue, which is not bad. 55 % increase in costs. Yes. And then 785 million of free cash flow. Yes. Which is essentially nothing. You know, a company like I ... you know, like I described earlier, that used to generate cash like water, now has no free cash flow. None.
And the other thing, you know, when you dig into the numbers, that was interesting, is depreciation. chips starts to cause damage. I think it was something like 6 million...6 billion for the quarter, compared to 4 billion maybe 3 months ago. I could be wrong, but not by much, but now all that capital that went on the balance sheet is starting to show up on the income statement, and it will be a burden on them for years.
Yes. If you were to think about how this is going to unfold over the next few months. I mean, Berry says it's the pinnacle. Or he says that maybe this is the top. I don't know how he can say that. I mean...well, of course you could say that. Right. I really don't know how you can claim that. Again , it's not like with subprime loans, when we had securitization data, where he and I had the same data.
There is no... He doesn't have a single indicator that you and I don't have . So if he had evidence that a price war had started between Anthropic, OpenAI, and Chinese models, I would say, "Okay, that's a very important indicator, and he's right." But otherwise I just think he's, you know, with all due respect to him, just pointing his finger at the sky , saying, "Well, let's try." And maybe he's right, but I don't think he has any data to point to .
I mean, look, as I said, the quality of bank loans is great, the employment data is very strong. Do you know what indicator you can point to right now that would say, "This is it"? What would you like to see if you were trying to determine the peak in this plan ? What would you need to see? Again, I need a price war.
I need a price war in the world of big language models. Until we see it, the market will grow. Until that happens, I think things will continue as they have been. OK. Steve Iceman, host of "The Real Ice Man Playbook." Steve, we always value your perspective. Thank you very much. It was nice to see you again.
We will be back soon. And if you like the show, don't forget to subscribe to the Prop G Markets YouTube channel at the link below. The show is supported by Vanguard. To all the financial advisors listening: let's talk about bonds. Making a profit from fixed income instruments is not easy. The bond market is huge and confusing, and let's be honest, many firms simply offer you a few flashy funds—and that's it , but not Vanguard.
Vanguard's bond line is built on institutional quality . And institutional quality is not just a slogan, it is a commitment to providing the investment solutions your clients deserve. This means access to over 80 bond funds, actively managed by a global team of over 200 industry specialists, analysts and traders.
While some companies like to single out one star portfolio manager, Vanguard takes a different approach. Here, it is believed that the most effective active strategies are born through collaboration and exchange of ideas between all members of the investment team. This means your clients benefit from the collective expertise of hundreds of investment professionals, not just one person's perspective.
So, if you’re committed to delivering consistent results for your clients year after year, check out our metrics at vanguard.com/audio. This is vanguard.com/audio. Any investment involves risk. Distributor—Vanguard Marketing Corporation. We return to the real estate markets. Due to the war with Iran and new tariffs, global supply chains are facing serious disruptions.
Yesterday, Iran and Oman agreed on a new route for ships passing through the Strait of Hormuz, moving closer to a deal, but Iran says it will not open the strait until the US lifts the blockade. This weekend, President Trump called off the “biggest attack since World War II” to give diplomacy another chance.
He warned that these talks are, and I quote, " the last chance to end the war." At the same time, the Trump administration has introduced new uncertainty into global trade. In late July, Trump invoked Section 301 of the 1974 Trade Act to impose sweeping new tariffs on 60 economies, covering 99% of US imports.
On Monday, 25 states filed lawsuits to block them, claiming the administration is using Section 301 to replace IEEPA tariffs. Of course, these are the same fees that the Supreme Court struck down earlier this year. So we wanted to better understand what's really going on and how it's affecting the global supply chain.
And who can tell you about it better than Ryan Petersen, CEO and founder of Flexport. Ryan, thank you for joining us. We haven't talked for a long time. I just want to make sure everyone understands your expertise. You run a logistics company. You have a great understanding of how supply chains move around the world in the freight industry.
And, of course, this is what is suffering greatly from tariffs and the war in Iran. And both of these situations either subside or flare up again. I can't understand what's really happening. What is happening? What is your opinion? Um, well, I think your summary is pretty apt. This, this is really confusing because the situation is constantly changing.
This, um, especially the war in Iran, is very difficult to understand. We have already had so many different peace agreements and " last chances" that it is difficult to understand how it will all unfold further. My predictive abilities in this matter are not that great. As for the tariffs, there is also a lot of uncertainty here, although I would say that there is a little more than there was before the introduction of these tariffs under Article 301.
Because, well, they will be challenged, as you noted, the state is filing lawsuits, and there are many other court cases. Section 301 tariffs have a better chance of surviving these lawsuits. There are a few more procedural points behind them. The IEEPA law is like a universal power, as interpreted by the president: he could just get off on the wrong foot, someone made him angry, and he imposes tariffs against the country, and Section 301 requires much more procedures, and they passed them.
They created, the Ministry of Commerce, excuse me, conducted a study to say that this or that country, this or that sector. It was much more detailed, much more thought out , let's say. So they are more likely to stand up. I always thought that IEEPA was shaky ground, most people thought so, so this is about the customs side of the issue.
It's a real whirlwind if you're an importer or exporter trying to cope with it all. So, how does this affect freight transportation and logistics? If you have a customs regime in place and you don't know what will happen tomorrow, and then it suddenly happens. Does this mean you start paying these fees right away?
Is there any grace period? How much does this confusion actually affect the movement of goods around the world? Yes, I mean it's about uncertainty and rapid change, and high stakes. I would probably separate these concepts. Last year we had both , didn't we? For example, customs rates are very high at a certain point.
I forgot, I think it was about 145% for China for a few weeks. Um, they still think quite highly of China, but of course there is a moment of high rates, but now they are much lower, even under Section 301, where you have about 10% or 12.5% depending on the country, and importers say that is a perfectly acceptable level.
That is, if everything remains at this level, stable and predictable, then they are completely satisfied. Um, last year was both : high stakes and a lot of unpredictability. So, in 2025, there were 52, no, 53 changes to the customs code in 52 weeks. This happened throughout the year. So it was complete chaos.
Some of these things were implemented without any warning . In fact, this was done almost without warning, as Article 122 expired at midnight on July 24, and this one came into effect at 12:01. And this was announced just a few hours before. So, there was very little time for planning. And this was manifested in the fact that many mistakes were made : companies filed incorrect declarations, paid incorrect amounts, and indicated incorrect duty rates.
And when it came to getting their money back, they had a hard time getting— well, they got their right to get their money back through a Supreme Court decision. And they have started coming in: Of the $166 billion in refunds owed under IEPA tariffs, $ 122 billion is currently in the process of being processed.
Um, I don't actually have data from the government on how much has been paid out, but a lot of that 122 billion is not being paid out or is being severely delayed, partly because the rules have been so difficult to follow that if you filed incorrectly, yes, you will get a refund, but you have to go through a whole legal process to first correct the record and then get a refund.
So people definitely face the consequences of disorganized data and incorrect submission of documents. I think this is a case, if I may briefly advertise my company, where we, being a technology company and having the databases to manage this, have gained a real advantage in helping companies do things right.
Actually, that's what I wanted to ask you about—the process of refunding duties. I mean, we know from the U.S. Customs Service reports that they were forced to provide to the courts because of all these lawsuits, that the administration probably paid out a hundred billion dollars in restitution after the Supreme Court ruling.
But, as you pointed out, that's exactly what you specialize in. You actually have a duty refund calculator to help people understand how to pay duties or get a refund . Um, and actually get those refunds. I mean, what stage of the return process are we at now? It sounds like it's a bit of chaos, but maybe that's too harsh a statement.
Well, I would still give the government credit on this issue. Overall, I think they did pretty well. They created a system called CAPE, which is a kind of application or module on top of the main Customs and Border Protection technology system. And they launched it in just two or three months, which is a truly remarkable achievement for government IT projects.
If we look back, the government's technology system for customs is called ACE, Automated Commercial Environment. It was a 10-15 year epic to develop this software. So for them to add a module as responsible and complex as reimbursing, you know, 100-166 billion in just a few months, in my opinion as a technologist, is pretty good.
I am somewhat impressed. The problems arise not so much because of customs. This happened because of the shortcuts that customs brokers and importers, that is, these companies, chose in an overly simplistic manner. So when you file, especially now when people are not getting reimbursed, you said it was a hundred billion.
I haven't seen these statistics, but yes, it sounds quite plausible . One hundred and twenty-two billion is the amount for which applications were submitted. And there are about forty- four or forty- two billion more lying around, not yet recoverable. This will be a separate, later process for these types of refunds.
But there are still twenty billion that were applied for but have not been returned. And most of that is steel and aluminum, where people made mistakes. There were these tariffs on steel and aluminum that were unique in the US customs system. Historically, you only needed to know three things to find out how much duty you owed.
This is the value of the goods, the country of origin and their HS code, i.e. classification by code. If you had these three things, you calculated the duty and knew how much you owed. Now, with the steel and aluminum tariffs that Trump imposed last year, you should already know that it's not enough to just know the product classification.
You need to know what percentage of the cost of this product is steel and aluminum. And you also need to know the country of smelting. Where this steel or aluminum was made, cast or smelted, depending on the production method. Oho. Companies didn't have this data, and if they did, they didn't know how to present it in a way that could be stored and verified.
Therefore, many simply falsified them. They said, "Well, okay, let's make it 25%." And there's a field there—this is a pretty technical point, so forgive me, viewers, if I bore you with this, but there's a field where you enter the amount of duty you owe. And many simply entered a number there and sent it to the government.
Maybe they calculated the duty amount correctly, but they had to allocate a percentage of aluminum; and you only have to pay duties on steel and aluminum on the value of the part of the product that is made of steel and aluminum. But these people were simply mistaken or looking for the path of least resistance.
So now when you want a refund, you can't get it on this return until you go back and sort out the whole process. This is a great example of how it's actually not that difficult, if you have a good database and you understand the rules, you can do it, but this is a new area. Many members of the customs brokerage community have still not been able to add this field to their software systems.
So you see the pain of an industry that just hasn't mastered the technology. But it's interesting to hear about all these complications, and yes, I guess all these details are technically boring, but it shows that we've created this incredible web of complexity, and I do n't even understand why. I mean , it's not to collect duties, it's to return them.
This is to undo all the complexity that was initially introduced. Yes, I think it's a very valid point that at some level this burden of compliance is as high as the burden of the tariffs themselves, you know, and that's exactly the case . The government has this idea that ... I think the direction that things are moving in— both in the United States and in Europe and other regions—is that governments want to have much more data about what's crossing borders.
They want to know, like in that example with the steel and aluminum, what is it made of? Not just a general classification of goods , but, for example, what percentage of this product is steel and aluminum? Where is this from? And we'll get to the point where they want to know this for every product.
What is this thing made of? Where does each of these subcomponents come from, because modern supply chains are incredibly interconnected, complex global structures. So it's very difficult for a company to keep track of all this because you're buying something... Yeah. You buy an object and insert it into a component, insert it into your product, but this component has subcomponents that come from different countries.
You don't know where, and now you have to trace it to first, second, third tier suppliers. And I mentioned that it's not just the United States. I mean, these rules are getting stricter. Europe now requires that when you import wood, you must provide GPS coordinates of where the tree grew. This is necessary to combat deforestation.
I think it's a noble cause, but now you have to show where the tree grew and then get a receipt from the truck driver to prove that this truck actually took the goods from that forest and brought them to the sawmill. And they want to see monitoring of this stage. This is where everything is heading.
And importers and exporters need to be proactive, using good database technology to track all of this. And you want this database to be networkable so that other people can contribute and add data to it, and so on. So I think from a Flexport perspective, we're positioning ourselves as: We believe we've made the right technology bets to help people manage this.
But as a free market advocate, I see the burden this places on companies. And there comes a point when you say, "Government, maybe you should charge me a couple of percent more in taxes, but don't make me do all this." Like, what are you trying to achieve? And it's not obvious. Yes, we are just getting in our own way, creating more and more complications for no real purpose, in my opinion .
I'll let you go soon, but before that I want to hear your thoughts on what's happening in the Strait of Hormuz. Is it closed? Is it open? If so, how much? What do we really know? As for container transportation, it can be considered practically closed. Now one or two ships pass through there per day, and at its peak there were almost 100.
So there is very, very little container traffic there. There are also significantly fewer tankers, although more of them pass through than container ships. They are probably more valuable to those economies. This, let's say, from my world—the world of container shipping—is a kind of "backwater," even though Jebel Ali is the ninth largest container port in the world.
This is in Dubai. Or it was before this war. This is actually because it is a hub for cargo transshipment; Many ships bring containers there, it can be considered a node in a hub-and-spoke network. And all this traffic has now moved to other places. They redirect these transshipments to another location.
Many things—to India. And, of course, they all bypass Africa. The Red Sea has been closed, in fact, since December 2023. Almost all container ships go around the southern tip of Africa rather than through the Red Sea, because of the Houthis, who are a kind of proxy group linked to Iran. So, I don't see an end to this .
I hope for a peace agreement, but even that probably won't provide enough confidence, because we've already seen these ups and downs. We already had peace agreements, and the next day the ships attacked again. So I don't know if this will calm the insurance markets. Insurance for one voyage costs about 10–15% of the ship's cost.
If you think about it, it's, you know...honestly, I don't even know how they evaluate such policies, because how do you know the probability that a ship will be hit by fire? Is it 10% or not? I mean, it's a pretty catastrophic, um, event, of course, and it's not just the loss of the ship, it's the entire cargo.
You could have a $200 million ship with a billion worth of goods that could sink. Ryan Petersen is the CEO and founder of Flexport. Ryan, I always appreciate your time. Yes. Hi, good to see you again, Ed. We will be back soon. And if you like the show, don't forget to subscribe to the Profiting Markets YouTube channel using the link below.
The show is backed by VCs. Public ticker for private technologies . For generations, American companies have moved the world forward through their ingenuity and determination. And for generations, ordinary Americans have been able to be part of that journey through perhaps the greatest innovation of all—the U.S.
stock market. It didn't matter if you were a factory worker in Detroit or a farmer in Omaha, anyone could own a piece of the great American companies. But now everything has changed. Today, our most innovative companies remain private rather than going public. The fact is that ordinary Americans are excluded from investment and left behind, while the chosen ones get all the benefits.
Until now. Introducing VCX, the public ticker for private technology , now available where you buy stocks. Fundrise's VCX gives everyone the opportunity to invest in the next generation of innovation, including companies leading the AI revolution, space exploration, defense technology, and more. Visit getvcx.com for more information.
This is getvcx.com. Please read the investment materials carefully before investing, including objectives, risks, fees and expenses. This and other information can be found in the fund's prospectus at getvcx.com. This is paid advertising. We return to the property markets. It's official. The artificial intelligence boom has become almost entirely dependent on OpenAI.
As we discussed, new Bloomberg reports confirm that roughly 70% of Microsoft's AI revenue last year came from just one company, and that company was , you guessed it, OpenAI. This would be alarming enough if it only concerned Microsoft, but it doesn't. Barclays estimates that 75% of Amazon's AI revenue came from OpenAI and Anthropic alone.
UBS estimates that 30% of Google's cloud revenue came from OpenAI and Anthropic alone. And if we were to convert this into revenue purely from AI, the figure would be about the same—75%. In other words, if OpenAI and Anthropic didn't exist, " big tech" would n't have an AI business at all. And the reason this is bad is because Big Tech has literally put everything on the line for one thing, and one thing only—AI.
Now you might say to me, “Who cares? After all, OpenAI and Anthropic exist, so everything is fine.” To which I will reply: "Yes, they exist, for now." Because do n't forget, OpenAI lost $21 billion last year. As for Anthropic, we don't know for sure, but our estimates suggest it's around $11 billion .
In other words, these companies survive only thanks to external subsidies. And who is this “someone else” now? Answer: big tech. The deeper you delve into the economics of AI, the more you realize it 's a house of cards . This does not mean that it will fall apart, but to prevent this from happening, no mistakes can be made.
Thank you for watching Profiting Markets from Profiting Media. If you enjoyed this episode, subscribe to our YouTube channel and join us tomorrow for our conversation with the one and only Professor Aswath Damodaran.
Comments 0