Summary of Key Points
On July 2nd, news that Meta planned to sell excess AI computing power, coupled with OpenAI's announcement of reduced AI operating costs, sparked concerns about overheated investment in AI. This led to a significant drop in the stock prices of leading Japanese and Korean memory chip companies (Samsung Electronics, SK Hynix, and Kioxia), with declines of over 5%, which in turn affected the overall Japanese and Korean stock markets. Industry insiders believe that this is not due to an actual surplus of AI computing power but rather a result of cloud providers selling off idle capacity to maintain cash flow when their AI investments have not generated sufficient returns. The International清算 Bank has warned that excessive investment in AI could pose financial risks, although it is expected that AI computing power investment will continue to grow rapidly in the long term, shifting from a focus on sheer spending to an emphasis on efficiency.
1. Why Did Korean Chip Stocks Suffer First When Meta Sold Its Computing Power?
Meta is a global leading cloud service provider that has acquired a large number of memory chips for its AI initiatives (AI requires storing vast amounts of data, and memory chips serve as the "data warehouses"). By announcing the sale of excess computing power, Meta is signaling to the market that it may not need as much hardware as initially anticipated due to potential underperformance in its AI operations.
Samsung, SK Hynix, and Kioxia are the dominant players in the global memory chip industry, with Meta being a major customer. When the market learned that Meta was reducing its demand for computing power, there was immediate concern that these companies' orders would decrease, leading to sharp drops in their stock prices: Samsung fell 6.5%, SK Hynix fell 7.7%, and Kioxia even dropped 12.6%, causing the Korean KOSPI index to plummet by nearly 5%.
2. Is AI Investment Really Overheated?
Industry experts argue that the issue is not a surplus of computing power but rather a mismatch between investment and returns. For example, Meta plans to invest $145 billion in AI this year, but if these investments do not generate profitable outcomes, cloud providers will have no choice but to sell off idle capacity to maintain cash flow.
OpenAI's recent technological advancements have reduced the cost of running AI by 50%, leading some to question the necessity of excessive hardware investment. However, long-term demand for AI computing power is still on the rise, according to TrendForce Research, with annual growth expected to remain in double digits from 2024 to 2028, peaking at 51% in 2026. Cloud providers will become more selective with their spending, aiming to achieve maximum efficiency with minimal investment.
3. Why Did the International Clearing Bank Warn About AI Investment Risks?
There are two main concerns:
1. Excessive Investment by Giants: Five major cloud providers, including Google and Microsoft, plan to invest over $1 trillion in AI between 2025 and 2026. If these investments fail (for example, if AI businesses do not generate expected profits), it could lead to financial instability, similar to the collapse of the internet bubble, resulting in losses for related companies and investors.
2. Risks from Private Lending: Private lending to the AI and IT sectors has increased fourfold over the past five years. If interest rates rise in the future, these companies may struggle to repay their loans, potentially leading to debt problems.
4. The Future of Memory Chip Companies: Cautious in the Short Term, Promising in the Long Term?
In the short term, cloud providers will be more frugal with their purchases of memory chips and will not stockpile them as extensively as before. However, TrendForce Research suggests that the impact is limited due to ongoing long-term demand.
For instance, Samsung and SK Hynix recently announced plans to invest 4755 trillion Korean won (about 24 trillion yuan) in semiconductor and AI data center projects in South Korea, indicating their continued confidence in the AI industry. After all, AI development is inseparable from memory chips, but companies will focus more on efficiency rather than reckless expansion.
Conclusion
The recent stock market downturn reflects market concerns about the uncertainty of short-term returns on AI investments. However, the long-term trend of AI development remains strong. Cloud providers are shifting from aggressive spending to a focus on efficiency, and memory chip companies will adapt accordingly. As long as AI continues to advance, their businesses will remain viable. Investors should not panic; short-term fluctuations are normal, and the key is to focus on long-term demand trends.