저자: JeremyGrantham · 외국인
요약 보기
미국 증시가 장기간 버블 구간에 있으며 AI 열풍이 2000년 닷컴 버블과 유사한 투기 광풍이라고 주장. CAPE 40(역사적 평균 17.6), 버핏지수 역사적 고점, 빅테크 AI 설비투자 과잉과 순환적 자금조달 등을 근거로 제시. 붕괴 시점은 명시하지 않으며 AI가 실제 작동하더라도 기업 이익률은 상승하지 않는다고 봄.
다듬은 글
Jeremy Grantham just said that the US stock market has been in a " bubble" zone for a long time. Sooner or later, this bubble will burst, and prices will return to their historical levels. Grantham is 87 years old. He has been managing capital at GMO since 1977. In the mid-1980s, he predicted the collapse of the Japanese market.
In January 2000, he opened 50% short positions in US stocks at the peak of the dot-com boom. And now he has published a 15- page article claiming that artificial intelligence is the biggest speculative craze since 2000. And he's not just talking about AI stocks, but the entire US market. His argument is that the bubble started to inflate even before ChatGPT appeared.
AI did not create this problem. He just laid on top of her. He describes it this way: AI hit the market , which was blowing up like a multi-stage rocket. The correction that began in 2022 has been halted, and the speculation that has arisen on top of an already overvalued market is what worries him now .
So, let's see what he says. So, GMO defines a bubble as a deviation of two standard deviations from the long-term real price trend of an asset. According to this indicator, the US market has been in a bubble zone for a long time. Now the numbers: the CAPE ratio, the cyclically adjusted price/ earnings ratio smoothed over 10 years, is 40.
Since 1880, the average CAPE value has been 17.6. At the peak of the “everything bubble” in 2021, it reached 38.6 — the second highest figure in history, and now it is already 40. The only time it has been higher is March 2000. And you can probably guess what happens next. There is also Buffett's indicator—the ratio of total US market capitalization to GDP, which is at a historic high.
The US now accounts for over 70% of the MSCI World Index. And as for individual stocks, Palantir is trading at a price that exceeds sales by more than 100 times, not earnings, but sales. Tesla has a price/ earnings ratio of over 300 with earnings falling 61 % year-over-year . These are not just meme promotions.
These are the main assets in popular investment funds . What does Grantham think about the future? He says: “When investor confidence eventually reaches its limit, the bursting of the AI bubble will lead to a serious disruption in the economy, falling profits and a significant decline in valuations.” He is also honest about the time frame.
The classic signs of a bubble peak are not yet apparent. He doesn't say it will all end tomorrow. He says that in his study of more than 300 bubbles in major developed equity markets, not one has escaped a return to the previous trend. Every one without exception. And he says this case will be no different.
So, Grantham wrote this article with Edward Chancellor . He is a financial historian who has dedicated his career to studying every major speculative mania imaginable . The South Seas Bubble, the British railway mania of the 1840s , the radio stocks of the 1920s and, of course, the dot-com era. In this article, he highlighted 10 signs that appear in all such bubbles.
All of these 10 characteristics are now inherent in the field of AI. So, the first sign is extravagant statements. During the early railroad mania, one newspaper proclaimed that the railroads would create one great family, speaking the same language, living by the same laws, and worshipping the same god.
Nikola Tesla said in 1904 that the entire Earth would turn into one huge brain. And now Sam Altman says that AI will bring incredible triumphs: fixing the climate, creating a space colony, discovering all the laws of physics. Mark Zuckerberg talks about a new era for humanity. So Chancellor's conclusion is this: never before have opinions about technology been so strong and polarized.
AI promises either heaven or hell. Does this remind you of anything? The second sign is easy money. The “everything bubble” of 2020–2021, which Charlie Munger called the most dramatic event in global financial history, emerged when rates were at zero . Of course, rates have risen since then, but compared to GDP growth or inflation, they remain historically low.
The third is investor optimism. Margin debt on the New York Stock Exchange has reached record levels. The share of U.S. household wealth invested in stocks is at an all-time high. Zero- day options on the S&P 500 now account for over 60% of all options trades. Investors are in euphoria, even though the University of Michigan's consumer sentiment index is near a historic low .
Fourth—new publications that advance technology. During the railway mania of 1845, 14 new weekly railway publications appeared. Wired was the main publication of the era of Internet mania. Now there is The Information, Rundown AI, and 100 new scientific papers on AI published online every day. Many of them are written by AI itself.
Fifth—the suggestion machine. When the ducks quack, feed them. In 2025, 60% of all venture capital in the US was invested in AI. OpenAI has gone from a valuation of $1 billion in 2015 to $ 750 billion today . Anthropic has grown from 4 billion in April 2022 to 61 billion now. I'm pretty sure it's bigger now.
And the wave of public IPOs has n't even arrived yet. But ducks quack very loudly. Sixth— suspension of normal valuation. The CAPE ratio is 40 versus the historical average of 17.6. As Sir John Templeton said, "This time it's different." These are the four most expensive words in the English language.
The seventh sign of a bubble, according to historical research, is immature technology. Gramp wrote that AI certainly remains an immature technology. Perhaps the biggest problem with modern large language models is that their errors or hallucinations look very convincing. Yann LeCun , one of the three scientists who laid the foundations of modern AI, argues that fundamentally new architectures are needed before AGI can emerge.
OpenAI co-founder Ilya Sutzkever said that the era of scaling is over. Terence Tao, considered one of the most prominent mathematicians of our time, wrote last December that he doubts the reach of true artificial general intelligence using current AI tools. The last three signs of a bubble are overinvestment, the “ bezel,” and the market clearing phase.
We will cover them in the next two sections, as they may prove to be the most important. So, this is where the numbers get really worrying. Amazon, Alphabet, Meta, and Microsoft alone spent nearly $ 300 billion on capital expenditures in 2025. This figure is expected to increase to 1.6% of US GDP in 2026.
Morgan Stanley estimates that total data center spending in the U.S. will reach $3 trillion— 10% of annual U.S. GDP—by 2029. Investment in AI already far exceeds the spending of the entire TMT sector in 1999 and 2000, and companies simply cannot stop. The article describes this as the "prisoner's dilemma." If they abandon the AI arms race, there is a risk that one of the competitors will emerge victorious.
Google CEO Sundar Pichai has publicly stated that the risk of underinvestment is much greater than the risk of overinvestment. So the very logic of competition guarantees collective overspending, but the expected cash flow never materializes. It is predicted that in 2025, OpenAI's revenue will be only $12 billion, and its operating loss will be $8 billion.
The company's own forecasts show losses doubling to 17 billion in 2026, and then to 35 billion in 2027. Total revenue across the entire AI industry is estimated at less than $50 billion, against a trillion dollars or more in investment. And then there's what the article calls the "bezel." This is Galbraith's word for the amount of undisclosed fraud that accumulates during a period of euphoria.
The Chancellor already sees clear signs of this phenomenon. Hyperscalers have stretched amortization schedules for AI chips from 3 years in 2020 to 5.5 years in 2025 . Google and Microsoft— both companies increased the term from 3 to 6 years. And this is despite the fact that the development of chip technology is likely to shorten their effective service life.
And then there is cyclical financing. Nvidia has committed $ 100 billion to OpenAI, which is spent on Nvidia's own products. Amazon invested in Anthropic, which uses Amazon Web Services. OpenAI has struck a deal with Microsoft's cloud platform, funneling Azure revenue back into purchasing Nvidia chips.
Although Microsoft owns 27% of OpenAI, the article compares this to the cyclical funding of the dot-com boom, and some say that the market cleanup has already begun. CoreWeave shares have fallen more than 50% from their peak. Oracle CDS spreads increased from 40 basis points to 140 in December. Shares fell 40%.
Meta's spending on AI has changed its position from a net cash balance of nearly $30 billion in 2023 to minus $7 billion today. And the scale of investment in AI is now causing DRAM prices to increase by 172% year-on-year and electricity prices to increase by 39% over 5 years. BlackRock expects this could lead to higher long-term interest rates.
Bubbles often burst after interest rates rise. If you found this video helpful, please consider becoming a member of the channel. You will receive my personal market analyses, the moves I actually make in the market, early access to certain videos, and exclusive member content. Just click the Join button right here and you can become a member to get these exclusive videos.
So, here is the argument that some may find the most interesting in the entire article. Because it's not just about whether the bubble will burst, it's about what AI actually means for the biggest companies, even if it works. Grantham's point is that even if AI lives up to all the optimists' expectations, it still won't significantly increase the aggregate profits of the companies that spend the most on it.
The reason is historical. In the late 1970s and 1980s, asset management companies rushed to buy minicomputers. The early players had a real advantage for maybe two or three years. Then the implementation became general. Everyone had to have them, just to stay competitive . Computers have changed the way work is done.
But the overall profitability has not increased anywhere. They have become a normal cost of doing business, like electricity or a telephone. Grantham says AI is moving along the same trajectory. As he puts it, this will not raise aggregate profits or overall profitability significantly above current levels.
Usually, Amazon has to use AI, otherwise it will lose to whoever uses it. Google must use it or it will lose search. Even the G7 companies are locked in an arms race from which they cannot escape. The moats they have dug for decades for protection are being depleted to fund military spending. In Grantham's words: "We have moved from a world of monopolies to a world of fierce competition , and we will remain there for years, and there will be blood in the streets." Don't forget, if you want more than just public videos, membership will give you exclusive access.
You'll see my moves , get early releases, and access to videos that no one else sees. Just click " join" below, and we'll see you there.
Comments 0