Summary of Key Points
U.S. storage stocks (such as Micron and Samsung) have experienced two consecutive days of significant declines, resulting in a total market value erosion of $340 billion. The direct triggers were two announcements from Meta (the parent company of Facebook): first, the plan to rent out its idle computing power for cloud computing services; second, Mark Zuckerberg's admission that progress with AI-powered intelligent agents is slow and internal reforms are not going smoothly. This has raised concerns in the market that tech giants may no longer continue to invest heavily in AI infrastructure, which could lead to a decline in demand for storage chips. However, there are differing opinions within the market—some are bearish (such as the prominent short seller Burry, who sold short Micron), while others are bullish (for example, NVIDIA CEO Jensen Huang, who described AI infrastructure as a decades-long project with sustained demand).
I. Sharp Decline in U.S. Storage Stocks: $340 Billion Lost in Two Days; Meta Announcements the "Trigger"
The decline was sudden and severe: Micron lost 15% in two days, and Samsung fared even worse (23%), with the four leading companies collectively losing $340 billion (approximately 2.4 trillion RMB). The direct reasons are as follows:
- Meta Plans to Rent Out Idle Computing Power: Meta invested heavily in building AI data centers, but now it has more computing power than it needs and wants to rent it out for profit. This raises concerns that if tech giants have excess capacity, they may no longer purchase new storage chips.
- Zuckerberg's Admissions: Zuckerberg acknowledged that the development of AI-powered intelligent agents is progressing slower than expected, and internal restructuring efforts have been inadequate. This further casts doubt on whether Meta's investment in AI infrastructure is as substantial as it seemed.
Storage stocks had previously seen strong gains because there was optimism that AI infrastructure would create a high demand for storage chips (such as memory and flash drives). With these updated expectations, stock prices naturally fell.
II. Behind Meta's Issues: Idle AI Infrastructure and Commercialization Hurdles
Meta's problems are not unique; they highlight two underlying concerns for tech giants investing in AI:
1. Underutilized Computing Power: Meta has built many AI servers but is now facing excess capacity, leading to the decision to rent out its resources.
2. Slow Progress in AI Development and Commercialization: The slow advancement of intelligent agents means that AI hasn't yet generated significant revenue, which may make giants hesitant to continue investing. The market fears that if these companies reduce their purchases, demand for storage chips will decline.
In short, the previous optimism about a sustained demand for storage chips due to AI infrastructure has faded with these developments.
III. Changes within the Storage Industry Affecting Stock Prices
Changes within the storage industry have also impacted stock prices:
- Declining HBM Profits: HBM (high-bandwidth memory) was in high demand for large AI models, but profits have declined due to supply issues. The cost of manufacturing HBM (which uses chips 4-5 times more expensive than standard DDR5 memory) and rising prices for regular DRAM have reduced HBM's profitability.
- Emerging Alternatives: As storage chip costs increase, companies are exploring cheaper solutions:
- Processor optimizations: Reducing the amount of temporary data needed during large model computations (e.g., using KV Cache technology to minimize memory usage).
- HBF (Hybrid Memory Fabric) solutions: Using cheaper flash drives to replace expensive memory/HBM components to lower costs.
These developments have dampened expectations for strong growth in the storage industry.
IV. Market Divisions: Bearish and Bullish Views
There are two opposing views on the future of storage stocks:
- Bearish View: Represented by prominent short sellers like Burry, who argue that Micron's stock price was excessively inflated and that its valuation, technology, and historical trends indicate a significant drop.
- Bullish View:
- CITIC Securities: Domestic orders for computing power are increasing, benefiting companies with relevant capabilities and production capacity.
- SemiAnalysis: Meta's demand for computing power has not slowed; it signed a deal for 5 gigawatts of capacity in the first half of the year (equivalent to 5 million servers) and is also involved in Musk's large-scale computing power leasing market (at prices 3-4 times higher than competitors).
- Jensen Huang: AI infrastructure is a long-term project with massive potential, involving electricity grids and the internet, ensuring sustained demand.
The core issue is whether short-term demand will slow down and whether there will be enough long-term demand to support storage chip sales.
V. Reversals in A-Shares and Japanese/Korean Storage Stocks: Emotional Recovery or Fundamental Support?
After the U.S. storage stock decline, A-share and Japanese/Korean stocks rebounded the following day:
- A-Shares: Baiwei Storage rose 3%, and Demingli rose 8%.
- Japan and Korea: SK Hynix, Samsung, and Kioxia reversed their declines of 5%-6% to gains of 4%-7%.
There are two possible reasons for this reversal:
1. Emotional Recovery: The U.S. market's sharp decline may have led to an oversold situation, attracting investors looking for opportunities.
2. Fundamental Differences: Domestic AI infrastructure demand is still growing (as indicated by CITIC Securities), and companies like Samsung and SK Hynix retain advantages in certain areas, maintaining positive long-term expectations.
These fluctuations show that while storage stocks are highly volatile in the short term, their long-term performance will depend on the actual growth of AI demand.
Conclusion
The recent decline in storage stocks was driven by changing market expectations. In the short term, prices may continue to fluctuate, but the key to future trends lies in whether AI infrastructure investment remains strong. Investors should focus on two indicators: whether tech giants are actually reducing their spending on AI infrastructure and whether AI applications (such as intelligent agents) will see rapid adoption. If these indicators are positive, storage stocks could still present opportunities.