AI Bubble Theory Heats Up and Semiconductor Divergence: Tech Sector at a Crossroads

Keywords: Artificial Intelligence, Bubble Theory, Semiconductor, TSMC, Memory Chips, JPMorgan, Internet Bubble

Introduction

In 2024, global capital markets experienced an unprecedented frenzy driven by the wave of artificial intelligence (AI). AI chip leaders like Nvidia saw their stock prices hit new highs repeatedly, tech stock valuation systems were redefined, and investors immersed themselves in the narrative of the "Fourth Industrial Revolution." However, any technological frenzy inevitably invites doubt and reflection. Recently, JPMorgan issued a rare warning: the current euphoria in the AI sector is showing precursors highly similar to those before the burst of the 2000 internet bubble. At the same time, significant divergence has emerged within the semiconductor sector – TSMC received a target upgrade from UBS, while memory chip leaders like Micron have experienced frequent volatility. Does this signal that the tech stock feast is about to end? Or is a structural adjustment brewing? This article will conduct an in-depth analysis from three dimensions: historical analogy, industry micro performance, and macro environment.

1. AI Bubble Theory: JPMorgan's Warning and Historical Mirror

1.1 Typical Characteristics of Bubble Precursors

JPMorgan's strategy team pointed out in a recent report that the concentration, valuation levels, and market sentiment of current AI concept stocks are highly similar to those during the 1999-2000 internet bubble period. Specific manifestations include: the proportion of AI-related stock trading volume to total market volume is at an all-time high; the P/E ratio of leading companies like Nvidia has exceeded 40 times, far above the historical average of the tech industry; retail funds are pouring in rapidly, leverage is rising, and analysts' earnings forecasts for AI companies show "optimism bias."

These phenomena do not exist in isolation. Looking back at the internet bubble, the market's almost blind pursuit of the ".com" concept led to high valuations for many companies lacking profitable business models, eventually crashing under the combined blow of rising interest rates and earnings disillusionment. Today, although the AI craze has a more solid technological foundation (e.g., the widespread application of large language models), some companies are still in the "pie in the sky" stage – revenue growth is highly dependent on AI narratives, but gross margins and cash flow have not improved correspondingly.

1.2 The Game Between Bubble and Structural Trend

It is worth noting that JPMorgan's warning does not completely deny the long-term value of AI. The bank acknowledges that the penetration rate of AI technology is still in its early stages, and its potential to boost productivity is real. However, "good technology" does not equal "good investment," especially when valuations have excessively discounted growth for several years. Historically, great inventions such as nuclear energy, airplanes, and personal computers all went through bubble phases, and only a few companies survived. The crowded trade in the current AI sector may precisely mean that investors lack precise discrimination of "true winners," instead treating all AI-related stocks equally.

The risk of this "sector-wide bubble" is particularly prominent in secondary markets: when funds collectively flee the AI sector, even companies with solid fundamentals (such as GPU designers and cloud infrastructure providers) cannot escape unscathed. This is the "liquidity stampede" risk that JPMorgan is concerned about.

2. Semiconductor Sector Divergence: TSMC's Stability vs. Memory Chip Fragility

2.1 TSMC: A "Safe Haven" Supported by Technology Moat

Under the shadow of AI bubble theory, TSMC, the leader of the semiconductor industry, received positive ratings from multiple investment banks including Morgan Stanley and UBS. UBS recently raised its target price for TSMC to NT$980, citing full capacity for its 3nm process and strong orders from AI chip clients like Nvidia and AMD. TSMC's uniqueness lies in that it is both an almost irreplaceable manufacturing link in the AI chip supply chain and avoids drastic fluctuations caused by changes in end demand – regardless of the ultimate adoption speed of AI applications, demand for advanced chip foundry will not disappear.

Additionally, TSMC's capital expenditure discipline and dividend policy provide a safety cushion for investors. The company recently announced it would maintain its annual capital expenditure plan unchanged while increasing the cash dividend ratio. This balancing strategy of "investing in the future while rewarding shareholders" appears particularly rational amid the AI frenzy, making it seen as a "stabilizing anchor in the bubble."

2.2 Memory Chips: A Source of Volatility Intertwining Cyclicality and Game

In stark contrast to TSMC is the memory chip sector. Stocks of memory giants such as Micron, Samsung Electronics, and SK Hynix have experienced significantly increased volatility recently. The reasons behind this are complex: on one hand, the explosive demand for high-bandwidth memory (HBM) from AI servers has made HBM the only memory chip category in short supply; on the other hand, traditional DRAM and NAND Flash are facing dual pressures of weak demand and high inventory.

For example, Micron reported better-than-expected AI-related revenue in its Q3 2024 earnings, but revenue from traditional memory business declined year-on-year, causing its stock price to first rise then fall after the earnings release. The market is starting to question: if AI enthusiasm cannot transmit to the general memory market, is the "good news" for memory chips just a local phenomenon? More crucially, Samsung and SK Hynix are also significantly expanding HBM production, which may lead to new oversupply risks in the future. This "ice and fire" structure makes the memory chip sector the most difficult to grasp in the current semiconductor industry.

2.3 Deep Logic of Divergence: From "General Rise" to "Selective Picks"

The divergence in the semiconductor sector is essentially an inevitable process of the market shifting from "concept-driven" to "earnings verification." In the early stage of the AI craze, almost all semiconductor companies rose because of "increased computing demand," and investors' buying logic was simple: as long as it touches AI, valuations can expand. However, as earnings season progresses, differences in profitability among companies are gradually exposed – TSMC earns stable excess profits through technology monopoly, while memory manufacturers suffer from commodity attributes and capacity cycles, with highly volatile profits.

This divergence also reflects the portfolio adjustments of institutional investors. A recent fund flow report from Goldman Sachs shows that hedge funds are reducing memory chip holdings while increasing positions in foundry and equipment stocks. This trend is further accelerating against the backdrop of JPMorgan's warning: when bubble concerns heat up, funds tend to flow toward sectors with high barriers and strong earnings certainty, rather than general pursuit.

3. History Doesn't Repeat Exactly, but Rhyming Patterns Warrant Caution

3.1 Similarities and Differences Between Current Environment and 2000

It must be acknowledged that simply equating the AI sector to the internet bubble is not entirely accurate. In 2000, many internet companies had not yet found clear business models, while today's AI companies have achieved substantive breakthroughs in revenue models (e.g., subscription-based API calls, cloud computing pay-per-use). Additionally, although the current global interest rate environment is relatively high, the Fed has signaled expectations of rate cuts, different from the tightening cycle background in 2000.

However, the similarities are equally significant: the market's absolute reliance on "superstar stocks." In 2000, a few companies like Microsoft, Cisco, and Intel accounted for most of the Nasdaq index's gains; today, a few AI leaders like Nvidia, TSMC, and Microsoft similarly dominate index performance. If this concentration reverses, it will have a huge impact on the entire market.

3.2 Risk of Zero-Sum Game: Fund Siphoning Effect

Another phenomenon to watch is the "fund siphoning effect." The continuous rise of the AI sector has attracted funds originally allocated to traditional tech, consumer, industrial, and other fields, significantly compressing valuations of non-AI sectors. If the AI bubble bursts, these drained funds may not immediately return but may choose safe havens (such as increasing bond or cash holdings), triggering broader asset price declines. This systemic risk is why JPMorgan believes that the "bubble theory" is by no means alarmist.

3.3 Investor Response Strategy: Stay Sober Amid Frenzy

Facing the heating up of the AI bubble theory and semiconductor divergence, how should investors respond? First, for high-valuation AI-related stocks, one should examine whether their earnings growth can match current P/E levels. For example, even if Nvidia achieves a 50% compound annual revenue growth rate over the next two years, its forward P/E remains above 30 times; once growth slows, the downside space for valuation is huge.

Second, consider "hidden champions" in non-AI applications of the semiconductor industry. For instance, chip demand in automotive electronics, industrial automation, and the Internet of Things has not surged simultaneously but has clear long-term growth trends. The valuations of such stocks are relatively reasonable and have low correlation with AI hotspots, helping to diversify bubble risk.

Finally, maintain liquidity management. The core of JPMorgan's warning is not that "AI companies will definitely crash," but that "current market sentiment is overly optimistic, and any negative news could trigger violent fluctuations." Therefore, investors should avoid excessive use of leverage and reserve some cash for counter-cyclical trading opportunities.

Conclusion: Bubble and Opportunity Coexist, Structure Matters More Than Direction

Artificial intelligence is undoubtedly a profound technological change, and its long-term impact may surpass that of the internet. However, the early stages of any technological change are inevitably mixed with speculation and illusion. JPMorgan's warning serves as an alarm: when the market begins to replace "company earnings" with "future of humanity" as the reason for investment, the cracks in the bubble have already begun to spread.

The divergence in the semiconductor sector provides a clearer signal: before the fervor subsides, true industry barriers and profitability will stand out. The stability of TSMC and the volatility of memory chips indicate that future investment opportunities will belong to companies that can convert AI technology into sustainable cash flow, not merely those relying on narratives.

Standing at the crossroads of 2024, investors need to make a choice: continue to follow the wave and enjoy the last euphoria of the bubble, or calmly assess risks and find certainty amid divergence? History tells us that there are no assets that rise forever, but there is always value that can cycle through cycles. For those who truly understand the industry, a bubble burst is not frightening – it is precisely the best time to separate gold from sand.

(Approximately 2,350 words)