Three stories of the AI ​​bubble: a Korean fried chicken shop owner, a haunted house owner, and zero GDP growth.

Warren Buffett (Berkshire Hathaway) has released new data, showing that he sold another $6.1 billion worth of stocks in the third quarter, bringing his cash reserves to a record high of $382 billion.

Of course, correspondingly, Buffett's returns are significantly lower than those of funds heavily invested in technology stocks, and even lower than the S&P 500 index.

What is he waiting for?

What is he waiting for?

Weekend mornings are a good time to watch the news at home.

I think the three recent news items are very representative of the signs of a mid-stage bubble.In financial history, every bubble burst has a few "landmark moments" that are constantly mentioned by later generations. It's like a joke, encapsulating all the madness of that era.

Just like in the movie *The Big Short*, hedge fund manager Michael Burry and his colleagues went to a prostitute and discovered that she had taken out loans to own 5 properties and naively believed that ARM interest rates would remain low for the next two years. Story 1: Huang Renxun sparked a "fried chicken" craze in South Korea. Huang Renxun's "Midas touch" also manifested in South Korean fried chicken. On Friday, photos and videos of Huang Renxun and Lee Jae-yong (Chairman of Samsung) eating fried chicken in South Korea went viral online: They even drank from a shared cup: The young woman next to them also became an internet sensation. (Of course, this was arranged; it's rumored that this is the daughter of Hyundai's daughter.) The most intense thing is that South Koreans have been frantically speculating on AI technology stocks in recent years. They've been buying Nvidia, TSMC, and even using leverage to buy QQQ (a Chinese tech stock), and some have even gone to Hong Kong to double their money on Hynix... Will the fried chicken that Huang Xiaoming ate be hyped up too? As a result, South Korean stock market investors began a frenzy, flocking to any stock even remotely related to "fried chicken," much like the surge in A-shares during Trump's presidential campaign when "Chuan Da Zhi Sheng" (a stock name) hit its daily limit. Although the Seoul-based fried chicken restaurant they ate at, Kkanbu Chicken, is not publicly listed, Kyochon F&B Co., a competing fried chicken chain, saw its stock price surge by as much as 20% on Friday. Cherrybro Co., a South Korean poultry processor, saw its stock price surge by 30%, hitting the daily limit, with trading volume 200 times higher than usual. Neuromeka Co., a company listed on the South Korean KOSDAQ that produces fried chicken robots, also saw its stock price skyrocket. Of course, Huang Renxun's reason for sharing a toast with Lee Jae-yong was primarily to secure upstream supply chain connections: South Korea is a crucial link in the memory industry, and memory prices have skyrocketed this year. (We'll delve deeper into why storage is so important today in the future.) Bloomberg commented that this event reflects how the internet boom and memes influence the South Korean stock market. In South Korea, risk-seeking traders chase short-term gains related to cultural, political, or economic hotspots, often ignoring fundamentals. The Financial Times stated that South Korean retail investors are "Asia's real gamblers," with data showing they are "avid ETF traders using triple leverage." In South Korea's ETF industry, leveraged assets account for a staggering 14.7%, compared to a global average of only 1.16%. Story Two: How a Haunted House Owner Joined the Data Center Construction Craze About an hour's drive northwest of Philadelphia, you'll find the ruins of a state medical facility on the banks of the Schuylkill River. Real estate developer Derek Strine has been running a haunted house there for years. He transformed the former medical facility, rife with scandals, into "Pennhurst Asylum," attracting tens of thousands of visitors every fall for meticulously planned scare events. However, Strine had a new idea for the land—he wanted to turn this nearly 130 acres of land, and this small town of only 7,000 people, into a world-class data center. He believed he could succeed because next to his haunted house was the Limerick nuclear power plant and its high-voltage power lines. In his own words, it was "untapped gold." He described his vision for a data center equipped with noise reduction technology and a closed-loop cooling system, and that it wouldn't draw groundwater. Of course, he had no experience in data center construction, but he was a master of speculation. He didn't need to actually build the data center; he just needed to act as a middleman. He called this plan "de-risk for the deep pocketed guys" because the greatest value lay in "achieving vertical integration of a hyperscale data center in less than a year." His plan can be divided into three steps: 1) He paid out of his own pocket to handle all the hardest and dirtiest work: smoothing over political resistance from the local government, obtaining complicated land planning permits, and (most importantly) getting approval for grid connection.

2) For this "queueing right", he has already paid $500,000 just to have the power company conduct an electricity study on him in 2026—the first step in his envisioned 500 megawatt grid-connected project.

3) Once he obtains this "birth certificate", he can resell this "shovel-ready" project, along with the nuclear power plant's "meal ticket", to Amazon, Google, or Meta at a high premium. This is a near-"get-rich-quick" scheme, reflecting the booming data center construction frenzy across the United States. Morgan Stanley says that $2.9 trillion will be allocated to data centers between 2025 and 2028. Of this, $1.3 trillion will be used for land, construction, and engineering costs; the remaining $1.6 trillion will be used to purchase GPUs from Nvidia and other companies. But GPUs are rapidly depreciating assets, so what can be done? Those smarter than us might come up with some dazzling structures to securitize assets that depreciate by 30% annually… Story Three: GDP growth – AI Capex = 0 The fried chicken frenzy and the haunted house speculation are micro-narratives. The third news item presents a disturbing macro-level picture. Harvard economist Jason Furman recently published a chilling macroeconomic statistic on X: In the first half of 2025, U.S. investment in "information processing equipment and software" (which you can essentially equate to capital expenditures on AI) contributed 92% to U.S. GDP growth during the same period. This means that if you exclude the expenditures related to this concept, the actual annual growth rate of the entire US economy—the global engine hailed as having “strong growth” and a “miraculous soft landing”—will be only 0.1% in the first half of 2025. The equation is frighteningly clear: US GDP growth - AI capital expenditure ≈ 0. The same pattern is observed in the US stock market, with the overall market gains highly concentrated in M7 and core technology stocks, while most other sectors have declined. However, if future layoffs due to AI and a significant underperformance of the overall US economy fall short of expectations, it will inevitably weaken the advertising revenue of tech companies, which remains the core business supporting their cash flow, since the actual profits from AI are still negligible. Meanwhile, the global spending of $2.9 trillion on data centers is an unprecedented infrastructure funded by private credit and high-yield bonds. And the demand upon which it is based—the AGI applications that can truly generate trillions in revenue—is, at least for now, still in the PowerPoint stage. The US economy hasn't grown across the board. It's just one industry (tech giants) frantically buying GPUs from another part of themselves (Nvidia), then including this "left-hand-to-right-hand" capital expenditure in GDP growth. What is Buffett waiting for? Buffett has been waiting, but the price of that "waiting" is a significant lag in his returns. Historically, when has Berkshire Hathaway's return significantly lagged behind the S&P 500 index? One was from 1994 to 1999. At that time, due to the dominance of internet stocks in the market, Berkshire's performance was severely hampered. Of course, after the bubble burst, Berkshire significantly outperformed the market between 2000 and 2003. Another was from 2004 to 2008. At that time, Buffett had also endured years of lagging behind. After the financial crisis, his defensive investment portfolio, combined with his previously accumulated cash reserves, directly "bought the dip in America" ​​(or rather, saved America), achieving huge excess returns. In the later stages of a bull market, when "growth" devolves into "speculation," Berkshire Hathaway's performance often underperforms the market. Personally, I still have great faith in Buffett's judgment because he possesses an unparalleled "information advantage." Because stock investments are merely the tip of the iceberg for Berkshire Hathaway; Berkshire's stock assets now account for less than 20% of its total net assets. Berkshire Hathaway has long been a diversified industrial conglomerate, comprehensively covering insurance, railroads, electricity, oil, automobiles, real estate, manufacturing, aviation, retail, and many other industries, comprising more than 180 companies. Although there are only 26 people at its Omaha headquarters, the entire group has nearly 400,000 employees. Berkshire Hathaway, to some extent, has become a "microcosm of the American economy" (excluding cutting-edge technology). All of this gives Buffett an unparalleled informational advantage, allowing him to gain a deep understanding of the American economy. For example, Berkshire Hathaway owns BNSF Railroad, the largest freight rail network in North America, and rail freight is a key indicator of economic activity. When data from the entire Berkshire Hathaway group is aggregated in Omaha, Buffett's mind acts like an "AI," constantly generating insights from massive amounts of data. Currently, stock market investors worldwide are frantically buying US stock ETFs (QQQ, SPY), especially South Korean investors who are heavily leveraging their investments. This situation is very similar to the "Nifty Fifty" ETF crash of 1973. (Due to space limitations, I'll write more about this history of the ETF bubble collapse in the future.) Of course, this article isn't advocating for an immediate sell-off. After all, not buying tech stocks today is like rowing against the current—you're definitely going to fall behind. Just a few days ago, there was news about a fund manager who was dismissed because his tech stock holdings were too small, resulting in significantly lower returns compared to his peers. Finally, I'd like to conclude with a story from *The Hedge Fund Chronicles* (authored by Barton Biggs, who founded Morgan Stanley Research and was a participant in the 2000 dot-com bubble before founding his own hedge fund): "The end of 1999 and the spring of 2000 were tough times because I prematurely downplayed technology stocks. In December 1999, I adjusted the proportion of technology stocks in my portfolio to..." My 15% allocation is less than half the total proportion of tech stocks in the S&P 500 and the EAFE (Australia, Europe, Far East) index. Perhaps this wouldn't be wrong at other times, but this was in 2000, when tech stocks soared amidst frenzy, and the tech bubble inflated significantly in the first six months of 2000. That was a painful spring. If the fastest-growing asset classes have such a low percentage in your portfolio, it's difficult to keep up with the index. Some investors were unhappy to see their assets lag behind, and I faced increasing criticism. Some of the points raised were quite valid. A Japanese friend told me an old Japanese saying: "The dancers are foolish, but the spectators are even more foolish." Others bluntly quoted Nietzsche: "Those who think others are foolish for dancing are simply unable to hear the music themselves." Hearing these comments, I felt even more like an idiot for not participating in this great tech ball. Even the young people in my office were somewhat dismissive of me. That summer, tech stocks soared to new heights, and the tech upstarts attending the conference were all veritable prodigies. Frank Quachon and Mary... Mick was their hero. Actually, Mary had warned that the situation was too crazy, but everyone thought she was joking. As entrepreneurs and executives, they were all confident and optimistic about the future. They talked about sustainable and disruptive technologies, growth curves, Gulfstream (airplanes), and "units." One "unit" is $100 million in net worth. "That guy is worth 5 units." This refers to a net worth of $500 million. Importantly, these people genuinely believe their boasts. I also find the wives who accompany their husbands to meetings quite amusing. They listen in, take notes, and wear designer jeans and high heels. Unlike the young, beautiful wives in Greenwich, they never talk about children or nannies; instead, they spread rumors about stocks. At cocktail parties, they endlessly discuss the size of mutual funds or tell anyone willing to listen how much money they've made buying and selling tech stocks. In 2000, at a heated investment discussion, Biggs expressed his views on bubbles and was ridiculed: "I said that human fear and greed have never changed since ancient times, and they drive the stock market to extremes. Old bubbles burst, new bubbles will form, and new bubbles are destined to burst as well. However, this sounded boring, and the audience was buzzing with whispers. The tech upstarts certainly weren't interested in the topic of bursting bubbles; they were preoccupied with the next IPO or business idea." At the end of his speech, Grassman said that the internet was the most important invention since printing and asked if I agreed. I said I disagreed. The internet is a groundbreaking technology, but there have been many inventions in the past century that are at least as important as the internet. "For example?" He pressed on. "For example, electricity, airplanes, telephones, computers… even air conditioning." This last addition might seem a bit silly. The audience erupted in commotion. He pressed on, asking if I really thought air conditioning was more important than the internet. Caught in a bind, I reluctantly continued arguing, saying that air conditioning had greatly improved the working and living conditions of hundreds of millions of people worldwide. Without air conditioning, many of the world's hottest cities might still be sparsely populated. There would be no Houston, Miami, Hong Kong, or Singapore. The southeast and southwest of the United States would stagnate. Without air conditioning, New York would become a sweatshop in the summer and fall. So, air conditioning is also an important invention, just as important as the internet in our lives. "Which would be more indispensable in your daily life? Air conditioning or the internet?" I asked. The debate moderator (an internet elite from San Francisco) then enthusiastically organized a vote among the audience. The question was: "Is an invention like air conditioning as important as the internet?" As a result, I lost face completely. The vote was 80 to 2, and one of the two people who supported me was my loyal wife. I still remember the pitying and embarrassed look on the face of the Morgan Stanley investment banker at my table when I returned to my seat. I heard whispers from another table: "He's outdated." As it turned out, my prediction was too early: tech stocks and the Dow Jones Industrial Average then soared for a full six months. In the heat of the moment, markets can always reach extremes you can't imagine. However, I consider that seminar I attended a milestone, marking the level of euphoria at the time, with even supposedly influential academics cheering it on. It seems to be animal nature. So, don't predict too early!