Last week, retail investors made a very typical "bottom-fishing gamble": injecting about $6.1 billion into the 3x long semiconductor ETF—the Direxion Daily Semiconductor Bull 3X ETF (SOXL)—in a single week, setting a record for the highest weekly capital inflow in the fund's history.

The result was not romantic. Chip stocks did not rebound; instead, they continued to fall.

SOXL fell another 10.8% this week, with a cumulative decline of over 43% since October; On Thursday, it plunged 20.5% in a single day, marking the largest single-day drop since June 23.

Behind this round of decline is not simply a technical correction, but rather the market beginning to seriously examine the investment returns of AI infrastructure.

| Why bottom-fish?

Since the beginning of this year, the semiconductor sector has performed exceptionally strongly. The iShares Semiconductor ETF (SOXX) once surged nearly 59% this year, and the Philadelphia Semiconductor Index repeatedly hit new highs.

So, when chip stocks pull back continuously, many investors' first reaction is: "If it drops too much, it should rebound." ”

But this time is different. Meta's Q3 report raised its capital expenditure guidance, and tech giants continued to expand their AI investments, sparking market concerns about "too much money being spent and returns coming too slowly."

The market no longer unconditionally rewards "spending money to build AI infrastructure," but instead asks: when will these investments turn into profits?

|Why did it lose so badly? 3x leverage is not simply multiplied by 3

SOXL aims to provide 3x returns on the daily price movements of the semiconductor index.

The key lies in the word "every day." It's not about simply magnifying a one-month or one-year index fluctuation by tripling, but about readjusting your position every day. In a market with continuous declines and increased volatility, such daily reset results in significant losses.

A more intuitive comparison:

A rough estimate of $6.1 billion in funds means a 43% drop could result in a book loss of about $2.6 billion. Of course, actual losses depend on each investor's specific buying timing, but this is enough to demonstrate the destructive power of leveraged tools during downturns.

What is the market afraid of?

The trigger for this round of sell-offs was Meta's earnings report. The company's capital expenditure in the third quarter reached approximately $34.9 billion, and the annual capital expenditure guidance floor was raised to $135 billion.

The market is not worried about the lack of AI demand, but rather:

In other words, the AI narrative remains, but the market is beginning to demand clearer paths to returns.

Kingtech's Perspective | Don't treat leveraged ETFs as stocks, and don't treat pullbacks as signals

For investors, this round of SOXL events offers three layers of insights:

1. Leveraged ETFs are not suitable for "buying more as prices fall" Products like SOXL are designed as short-term trading tools, not long-term allocation tools. During consecutive declines, daily resets accelerate principal loss, and rebounds may not compensate for earlier losses.

2. AI sector hasn't crashed, but valuations are being repriced. Semiconductors, storage, optical modules, AI servers, and other sectors still benefit from computing power development, but the market no longer gives all related stocks the same premium. Next, more attention will be paid to orders, cash flow, and capital expenditure fulfillment.

3. For A-share and Hong Kong investors, focus on the industry chain delivering results. Fluctuations in the US AI chain will affect global valuation sentiment.

Domestic sectors such as computing power, semiconductor equipment, storage, and optical communications need to be more strictly selected: prioritizing companies with orders, production capacity, and customer validation, rather than simply following conceptual fluctuations.