원문 보기
Two of the most successful investors in existence observed the exact same market and came to opposite conclusions. One of them is Paul Tudtor Jones, the trader famous for predicting the 1987 crash. He says it's 1999 and told CNBC he was sure that at some point he would be shorting the entire year. He says it's 1997 and that the real peak might not arrive until 2033.
They can't both be right. So today I'm going to put their arguments side by side , check the numbers, and show you the metric that I believe decides who wins. Here's a clue: it's not the stock price. Let's start with the bear because it has the most terrifying history behind it . Quick context. Jones has been managing money for more than four decades and is best known for predicting Black Monday in 1987.
So when he gives so many details about a single year, Wall Street takes his word for it. Last week on CNBC, Jones was decidedly pessimistic about US stocks in 2027. He said it will be a very difficult year and described supply and demand as the worst since 2002. This is the logic. For four years, buybacks and mergers have been reducing the number of shares on the market.
Now, that is being reversed. SpaceX just raised a record $75 billion in its initial public offering, and Anthropic and Open AAI are expected to follow. That's an avalanche of new coup actions. The interviewer asked him exactly that. SpaceX, OpenAI, Anthropic: Will there be a displacement effect? Will a market void be created in 2020?
I'm sure that at some point in 2027 I'll be short of money all year. No doubt. You'll be short all year round. Well, again, if we go back to the year 2000, yes, you remember it. If we go back to the year 2000, there were all those issues, all those IPOs in 1998 and 1999, and actually 2000, 2001 and 2002 were the years in which it was easiest to be short.
His second argument is about who is actually doing the work. At the beginning of October, the S&P 500 had risen about 13.8% for the year. It sounds healthy, but the average stock in the index had only risen 3.5%. 220 members, 44% of the index, dropped in the year. And on a day when the index reached a record high, only 15 of its 504 stocks were at all-time highs.
Nvidia alone accounted for about 2/3 of the group's value. That's the pattern Jones remembers from 1999, when a small group of internet and technology names were doing most of the work and much of the market was going nowhere. Zooming out, the 10 largest stocks now account for about 40% of the S&P 500.
At the peak of the bubble, that figure was 27%. According to this measure, the market is more unbalanced today than it was in 2000. Then there is the cost of money. Earlier this month, the yield on 10-year Treasury bonds reached 5.31%, the highest since 2002. Stocks are near all-time highs, while bonds are paying more than 5%.
That's not a normal combination. Jones also maintains a long position in crude oil, betting on an additional $20 or $25 per barrel due to the conflict with Iran, and warned that stocks would hate such a rise. He also pointed to the federal deficit. His view is that Washington only acts when the markets force it to , as in 2008.
When asked if we were already on that path, he estimated the odds at 60 to 40. And he's not alone. Ray Dallio calls AI a classic bubble. And its trigger is the same one that Jones observes. Raising interest rates pushes those with profits in paper to sell. Even within the AI sector, the tone is changing.
Broadcom CEO Hawk Tan, whose company manufactures custom chips for this development, said that many investors have no idea what they are investing in. He compared it to a casino. Now, the other side, Tom Lee's argument, begins with psychology . In his words, I have never seen a bubble when everyone else believes it is a bubble.
When everyone is preparing for a crisis, it's difficult to achieve the euphoria that a real bubble needs. Then it moves on to the scale. It estimates the development of the Internet, mobile networks and mobile phones at about 35 trillion dollars in current money. He estimates the AI infrastructure at around $30 trillion.
And he maintains that we are still at the beginning of that spending. It also says that around $7 trillion of new liquid wealth is created each year to help pay for it. Its timeline comes from wireless technology. 25 years of exponential growth with the stock reaching its peak after 10 years. When applying this to AI, he reaches a peak around 2033.
But this is the part of Lee's argument that many overlook. Their roadmap is Cisco, the Nvidia of the dot era. Cisco did not reach its peak until March 2000. Before that, it fell by 50%. Twice, and after the peak , he lost almost 90%. Even optimists warn of heartbreaking falls on the way up. There is also a big difference.
Cisco didn't become the world's most valuable company until 2000. Nvidia is now number one with about $5.8 trillion. This cycle begins from a much higher base . And it's not just Cisco. In 1999, Amazon was a 5-year-old company and the darling of the dot-com bubble . In the following 2 years, the shares lost more than 90% of their value.
The business survived and became one of the most valuable companies in the world. But anyone who bought at the top waited years just to be evenly matched. That's the uncomfortable truth about bubbles. You can be completely right about technology and still lose a lot of money with stocks. If you put both opinions on the same graph, something interesting happens.
Jones believes we are at the end of the cycle, around 1999. Lee believes we are in the middle of the cycle, around 1997. But they are closer than the headlines suggest. Jones does not predict a drop tomorrow. He said the market could still rise until the end of 2026. His short position is for 2027, and Lee's own Cisco roadmap includes multiple 50% drops.
Ask him if 2027 could be violent and his own framework says yes . So the real disagreement is not about ups or downs, but about how long the cycle lasts. Which brings me to the figure that I believe decides it. Whether customers who buy AI hardware are actually making money from it . In 2000, Cisco's biggest customers were young internet companies and carriers.
Many of them bought equipment with borrowed money. When their revenue stopped coming in, orders stopped and Cisco went bankrupt. Today, Nvidia's biggest customers are some of the most profitable companies on the planet. Alphabet, Amazon , Microsoft, and Meta planned more than $700 billion in capital spending this year.
The question is whether that expenditure pays off. This week we obtained a real piece of data. As we covered in our last episode, SpaceX said its AI computing is being leased at rising prices, recouping new investments in less than a year. This is what a healthy cycle looks like. But SpaceX also invested about $2.18 in AI capex for every dollar of revenue last quarter.
That gap needs to be closed. So this is my opinion. The ingredients of a bubble are clearly here. Extreme concentration, a wave of mega-IPOs and 5% bonds with stocks near their all-time highs. Jones is right that 2027 will have supply problems. But bubbles burst when customers stop making money. And we have n't reached that point yet.
So I would put it this way. The question is not whether there will be a big fall. History says there will be one. The question is whether AI customers are still earning more than they spend. Three things to keep in mind. Big gains in the cloud in the next two weeks. Cloud growth versus capital expenditure.
The IPO schedule, especially Anthropic and Open AI, and the 10-year return if it exceeds 5.3 %. If this breakdown has helped you, subscribe. I will be following both sides of this operation closely. As always, this is not financial advice. Do your own research and see you next time .
Comments 0