When buying winners and selling losers, momentum trading is taken to the extreme, and the stock market frenzy driven by AI reaches the "wake-up moment."

The most classic momentum trading strategy in the stock market, which is to buy winners and sell losers using various technical indicators or institutionally customized quantitative models, has been popular in the US stock market at the beginning of this week's trading. Now, amidst a temporary pause in Iran's war, stronger-than-expected US non-farm payroll data, and rising risk appetite, the global stock market bull market revolving around AI computing power trading is becoming increasingly hot, adding fuel to this trading strategy. The core of the so-called "momentum trading" is to buy assets/stocks that have performed the strongest in the past period, and sell or short assets/stocks that have performed the weakest, also known as "buy winners, sell losers." By the end of Friday's trading in the US stock market, strong employment data and the trading frenzy around AI computing power drove both the S&P 500 and the Nasdaq 100 to new all-time high levels on Friday, with the S&P 500 recording six consecutive weeks of gains. The Philadelphia Semiconductor Index, a benchmark index of global chip stocks that include leaders in the global AI computing industry chain such as NVIDIA, AMD, Intel, and Micron, surged more than 5.5% on Friday and rose 11% this week, also notching six consecutive weeks of gains, skyrocketing nearly 250% from its low point in April, with its valuation reaching a historical high based on market-to-sales ratio. In addition, similar risk preference trading strategies are proving effective in various risk assets outside the stock market, from junk bonds to cryptocurrencies. In the US stock market, a momentum index closed at nearly its highest level since the global financial crisis; at the same time, the S&P 500 hit a new high again, the Philadelphia Semiconductor Index surged significantly in five trading days, and international oil prices fell as the market expected the end of the Middle East geopolitical conflict. However, the global stock market frenzy around AI computing power trading seems to have entered a phase of rapidly increasing sentiment and positions. Some Wall Street analysts have even warned that when oil prices fall, geopolitical tensions ease, corporate profits exceed expectations, and concerns about Federal Reserve tightening subside simultaneously, the market is prone to enter a downward adjustment trajectory of "good news already priced in" or even a bear market pullback. Strategists from the Wall Street financial giant Barclays pointed out that the momentum indicator has reached extreme levels that have often forewarned sharp sell-offs in the past. The trading department of Goldman Sachs wrote this week that based on primary brokerage data, the valuations of the highest momentum stocks have been overly stretched, and market positions are at some of the highest levels in recent years. Momentum trading in the context of Wall Street institutions is not about relying on a few technical indicators, but using a combination of relative strength, performance over the past 6-12 months, excluding recent one-month reversal effects, volatility adjustments, industry neutrality, risk models, crowding, fund flows, and volume signals to create a customized quantitative model to guide momentum trading strategies or momentum factors. Overall, the AI-driven bull market in the US and global stock markets has not been disproven, but in the short term, it has transitioned from an "earnings upgrade rally" to a "momentum trading overheating-driven pullback phase." In this phase, the odds of continued chasing winners in the AI stocks significantly decrease, and volatility increases; a more reasonable strategy is not to deny the AI computing power investment theme but to reduce crowding exposure, enhance quality selection, focus on profit realization and cash flow, and use short-term downside protection to hedge against the risk of sharp declines in semiconductor/AI leaders. The investment frenzy around AI computing power is pushing momentum trading to its limits Historical data suggests that momentum trading is becoming increasingly extreme, implying that the bull market in US stocks driven by the AI investment frenzy has entered a phase of "overheating/crowding/high fragility"; however, it more accurately signals a period of downward pullback or increased risk of sharp rotations, rather than the inevitable end of the bull market. In particular, extreme momentum trading combined with stock market overvaluation typically indicates that the market has overpriced the expectations of fundamental expansion when fundamental data or macroeconomic conditions significantly deviate from expectations (such as profit downgrades, major changes in interest rate expectations, rising financing costs around yield curve), bull markets are more likely to experience periods of pullbacks. The dominance of trend-following trading reflects the strong resilience of the US economy and the market's firm bet on the unprecedented wave of AI, especially the leading chip companies closely associated with AI computing infrastructure. For some seasoned strategists on Wall Street, this trading strategy has started to exhibit self-reinforcing characteristics the rise in a small group of tech giants attracting more funds into passive ETFs holding the same stocks. Even those participating in this rally acknowledge that, currently, the fundamentals are still supporting this investment theme. Mark Hawtin, Global Stocks Trading Director at Liontrust Asset Management, said, "Valuation factors have temporarily been thrown out the window because, undoubtedly, there is a sustained shortage of AI computing capacity." The asset management and advisory firm manages about 20 billion (approximately $27 billion) in assets. "When will this change? That is a million-dollar question, and it is extremely difficult to answer because I believe liquidity is very abundant." In his view, the supply-demand imbalance in the AI computing infrastructure will eventually significantly ease at some point, leading to lower high corporate spending in areas such as semiconductors. In a market driven mainly by momentum trading, being too early in betting on rotations can be as damaging as being wrong. Three cloud computing super giants, Microsoft, Google, and Amazon, delivered impressive performance reports on the same night, highlighting the unexpectedly rapid growth of their cloud computing businesses benefiting from the AI wave. The latest research report from Morgan Stanley's analyst team shows that with these giants significantly increasing their spending on AI computing power, the five mega tech giants - Amazon, Google, Meta, Microsoft, and Oracle - are expected to have a combined capital expenditure of around $800 billion by 2026, potentially surpassing $1.1 trillion in 2027, a revision higher from the previous forecast of $950 billion. Morgan Stanley's analysts emphasize that the core logic behind these massive capital investments lies in building capacity based on AI computing resources first, and then deriving revenue and ROIC through scalable business operations. The surge in cloud computing backlogs is the most direct evidence that this logic is working, and the unexpectedly rapid growth of the cloud computing business of these giants is prompting Wall Street to reprice the commercial returns of AI. For the stock market, a series of positive news has kept the animal spirits high this week. The US saw better-than-expected job growth in April, while overall positive progress in the Iran agreement significantly lowered oil prices. Among the so-called factor combinations tracked by quantitative investors, momentum is a prominent chameleon that adapts to any asset that is on the rise in the market, usually measured over a period of about a year. It has a history of total collapses when market leadership suddenly switches, such as during the global financial crisis in 2008 and when the COVID-19 vaccines were first rolled out in 2020. Arun Jain, head of the quantitative strategy team at J.P. Morgan, and his team warned this week in a report that retail investors are showing irrational behavior, with their buying intensity returning to levels seen just before the geopolitical conflict in the Middle East started at the end of February. Stocks related to AI computing infrastructure, including AI GPU/ASIC, data center power chain, HBM, advanced packaging, and storage chips, are still their favorite investment themes. For Hawtin from Liontrust, the strength of momentum trading strategies in recent years has reshaped his team's ways of operating in the market. They rely on technical indicators to follow trading trends, even if other indicators may give opposing signals. "For some semiconductor stocks, it does feel a bit like the peak of a so-called casino moment," he said. "From a long-term investment perspective, these stocks are currently trading at levels that are not rational, but as we know, markets can sustain irrational trades for a period of time." The S&P 500 hitting new highs again, the chip index rising 11% in five trading days, AI winner stocks rising over 50% from the lows in late March, and the momentum index nearing its highest level since the global financial crisis; Barclays believes that such extreme momentum levels have often signaled subsequent sell-offs in history, and the Goldman Sachs trading desk also warned of stretched valuations of high momentum stocks and near-record high positions. However, as momentum trading becomes increasingly extreme, it indeed highlights that the bull market in US stocks driven by the AI investment frenzy has entered a phase of "overheating/crowding/high fragility"; but it more accurately signals a period of pullback or increased risk of sharp rotations, rather than the inevitable end of the bull market. From a quantitative perspective, the factor that momentum strategies fear the most is not "poor fundamentals" but a sudden shift in market leadership. When funds are highly concentrated in AI chips, data center CPUs, storage, data center power, and a few large tech stocks, the rise reinforces itself: rising stock prices lead to increased index weights, which in turn attract passive funds to continue buying winners, further pushing up their stock prices. However, this structure also implies that once there are fluctuations in oil prices/geopolitical tensions, a cooling of AI capital expenditure expectations, profit guidance falling short of expectations, or a shift of funds from AI winners to laggard sectors, crowded trades will be forced to deleverage, and the speed of the pullback is often faster than a normal valuation correction. This does not equate to an immediate burst of the "AI bubble" or an immediate end to the bull market dominated by the AI frenzy. For example, the ongoing rise in US stocks still has fundamental support: approximately 85% of S&P 500 companies have exceeded earnings expectations this earnings season, the shortage of AI computing capacity is real, the increasing capital spending by cloud giants, HBM/storage, AI GPUs, and data center power chains are delivering profits. However, the market has already priced in a "perfect combination" of "continuing AI demand shortage, sustained profit upgrades, and sustained macro-friendliness"; when valuations, positions, sentiment, and technical indicators are all at high levels, even without a fundamental reversal, a mere cooling of expectations can trigger a 10% range pullback or a 25%-30% technical correction in the semiconductor sector.