The AI rally hasn't ended; it has entered a phase of distinguishing between companies that will recover their costs and those that will bear them. [Photo=Hanmi Ilbo Graphic]The stock market this week hasn't abandoned AI, but rather has begun to re-examine the AI sector with more scrutiny. 

While Nvidia, Micron, SanDisk, Intel, AMD, and semiconductor equipment stocks were shaken simultaneously, Meta and certain large-cap tech stocks moved in a different direction. Even within the same AI theme, stock prices have diverged based on their potential for monetization and their cost burdens.

The question for Stock Radar last week was this:

“Has the AI semiconductor rally ended, or has the memory supplier taken the lead?”

The answer last week was closer to the latter. Micron’s earnings confirmed the demand for and pricing power of AI memory, and the market perceived that within the AI value chain, profits were shifting more toward suppliers than consumers.

The answer this week has become a step more complex. 

While the leadership of memory suppliers remains intact, a signal has emerged that the market will no longer pay any price for that leadership. The fact that AI demand exists is one thing; how much of that demand can be reflected in current stock prices is quite another.

The name of this week’s trend is ‘AI sector differentiation.’

News that Meta could potentially sell its surplus AI computing resources to third parties was interpreted as positive for the company. This is because it could create a new channel to recoup massive AI investment costs. 

However, the same news was read as a negative for semiconductor suppliers. The market seized on the phrase “surplus computing resources” as a potential indicator of overinvestment in AI infrastructure.

This interpretation is not the final verdict. The ability to sell AI computing resources externally could, in fact, indicate that demand for AI computation remains robust. 

If Meta generates profit from a cloud-based AI computing business, it will require even more chips and data center investments in the long run. Therefore, this correction is less about the collapse of AI demand and more about a conflict in how that demand is interpreted.

The question for this week is:

“Is the semiconductor correction a signal of collapsing AI demand, or an intermediate adjustment driven by pricing and supply-demand dynamics?”

The answer remains closer to the latter. This week’s correction is the result of price pressure, leveraged positioning, and the overinvestment narrative coming into play, rather than a sudden collapse in fundamentals. 

However, the market no longer views AI as a single theme. It has begun to evaluate Nvidia, memory, equipment, data centers, cloud services, and big tech monetization models separately.

The conclusion in one sentence is this:

“The AI rally hasn't ended; it has entered a phase of distinguishing between companies that can recover their costs and those that will bear them.”

There are four checkpoints for next week:

First, we must confirm whether the decline in memory stocks like Micron and SanDisk is stabilizing. 

Second, we need to see if the correction in Nvidia continues to spread to equipment stocks and Korean memory manufacturers. 

Third, we must check whether the debate over Meta’s surplus computing resources solidifies into a narrative of slowing AI demand or is reinterpreted as a new monetization model. 

Fourth, we need to confirm whether earnings expectations for Samsung Electronics and SK Hynix hold firm despite the correction in US semiconductor stocks.

※ This material is for reference only and summarizes market trends; it is not a recommendation to buy or sell any specific financial instrument.