South Korea’s stock market has seen heightened volatility recently, sparking debate over whether the AI semiconductor rally is cooling down. According to Korean media reports, from July 1 to 16, foreign investors made net sales of about KRW 12.1 trillion in the KOSPI market and approximately KRW 338.1 billion in the KOSDAQ market. During the same period, the KOSPI index fell sharply from its late-June peak, significantly impacting market sentiment.
Over the past year, one of the biggest drivers of gains in the Korean stock market has been the AI semiconductor supply chain. With global investment in artificial intelligence surging, Korean companies have leveraged their strengths in memory chips to become major beneficiaries. Among them, SK Hynix has become a key player in NVIDIA’s AI GPU supply chain thanks to its High Bandwidth Memory (HBM) technology, attracting significant global capital and attention to its stock performance.
As a result, when foreign investors significantly reduced their positions in Korean equities, the market’s focus naturally shifted to the AI semiconductor sector. However, a closer look at capital flows reveals that this adjustment is not simply a flight from AI, but rather a reallocation of assets within the AI supply chain.
On one hand, foreign investors are selling Korean stocks; on the other, they continue to buy U.S. Philadelphia Semiconductor Index ETFs and Nasdaq-related products. This indicates that global capital remains interested in the long-term AI trend, but is adjusting its portfolio structure.
Why Is SK Hynix a Focus of Foreign Investor Adjustments?
SK Hynix is one of the core companies in the spotlight. Over the past year, the main investment thesis for SK Hynix has centered on HBM. As generative AI rapidly develops, large language models require greater data processing power, and improvements in GPU performance have driven demand for high-speed memory.
Traditional DRAM memory primarily serves the PC, smartphone, and general server markets. In contrast, HBM is designed for AI accelerators, using multi-layer chip stacking and high-speed data transfer to boost GPU efficiency in AI training and inference.
This shift has transformed SK Hynix from a traditional memory-cycle company into an AI infrastructure supplier. However, because the market now fully recognizes HBM’s value, SK Hynix has become one of the most heavily traded AI stocks. When volatility hits, these popular assets are often the first to see capital adjustments.
Foreign investors reducing their holdings in SK Hynix does not necessarily signal a lack of confidence in HBM demand. Instead, it may reflect several factors:
- AI semiconductor stocks have seen substantial gains, prompting some investors to lock in profits;
- As the market focuses on the expansion of HBM supply, there are questions about future pricing and profit margins;
- Samsung Electronics and Micron Technology are also ramping up HBM efforts to gain market share. As the competitive landscape evolves, investors must reassess each company’s position in the AI memory cycle.
What Does It Mean When Foreign Investors Sell Korean Stocks but Keep Buying AI Assets?
A notable trend in recent capital flows is that foreign investors have not completely exited AI. Data shows that while selling Korean stocks, foreign investors have also made net purchases of Korean ETFs, including products like KODEX 200 and KODEX Leveraged ETFs. At the same time, they are buying U.S. Philadelphia Semiconductor Index ETFs and Nasdaq 100-related ETFs. This shift suggests that capital is moving from single-stock investments toward a more diversified allocation across the AI supply chain.
Over the past year, the market’s AI investment playbook has been relatively concentrated: buy NVIDIA, focus on SK Hynix, and bet on HBM.
But as the AI industry matures, investors are seeking broader opportunities. The AI supply chain is expanding from single-chip competition to a full-scale infrastructure race.
This includes: AI GPUs, HBM memory, high-speed interconnects, data centers, semiconductor equipment, and power infrastructure.
Therefore, foreign investors reducing their holdings in Korean stocks while increasing allocations to U.S. semiconductor ETFs does not mean the AI trade is over—it likely signals that the market is entering its next phase.
From SK Hynix to the Global AI Supply Chain: Capital Is Seeking New Opportunities
The investment logic for AI is evolving. In the early stage, the market focused on who could deliver the most powerful computing.
NVIDIA was the core player in this phase. As GPU demand exploded, HBM emerged as the new bottleneck, bringing SK Hynix, Micron, and other memory companies into the spotlight.
In the next phase, investors are shifting their attention to whether AI infrastructure can support ongoing compute growth. This elevates the importance of high-speed networks, optical interconnects, data centers, and energy systems.
For example, Broadcom continues to benefit from AI networking chips and custom ASICs; Marvell participates in AI infrastructure through high-speed interconnect and data center network solutions; and data center operators are gaining attention as AI server deployments increase. The AI industry is forming a more complex value chain. In the future, the market may not reward just a single AI leader, but will seek out key links across the entire ecosystem.
Korean semiconductor firms remain vital parts of the AI supply chain, but capital allocation strategies are changing.
Has the HBM Cycle Peaked?
This is one of the market’s most pressing questions. Historically, the memory industry is known for its cyclical nature. When demand rises, companies expand capacity, prices climb, and profits improve; but as supply increases, the market can enter a price correction phase.
Whether HBM will repeat the traditional memory cycle is a key concern for investors.
In the short term, as Samsung, Micron, and others ramp up investment, HBM supply capacity is increasing. If supply grows faster than demand in the future, prices may come under pressure.
However, unlike traditional DRAM, HBM has a much higher technical barrier. It requires not only advanced memory chip manufacturing, but also cutting-edge packaging, supply chain coordination, and deep collaboration with GPU companies. As a result, competition in the HBM market is not just about capacity, but about comprehensive technological capabilities. As long as AI data center demand continues to grow, high-performance memory will retain strong strategic value.
The real question for the market is not whether HBM’s growth is over, but who can maintain leadership in the next phase of competition.
AI Semiconductor Investment Enters a Supply Chain Restructuring Phase
Recent adjustments in the Korean stock market reflect a trend: AI investment is shifting from a rapid growth phase to a period of structural realignment.
The market primarily trades on AI themes and compute demand.
The focus is on core supply chain companies.
Investors are starting to seek long-term value at the infrastructure level.
In the future, competition in the AI industry may resemble traditional tech infrastructure development.
Owning advanced chips is just the beginning. The true enablers of large-scale AI adoption will include:
- Sufficient HBM supply;
- High-speed interconnects;
- Adequate data centers;
- Reliable energy infrastructure.
This explains why the market’s attention has recently shifted to AI data centers, power infrastructure, and semiconductor equipment companies.
AI is evolving from a software trend to a global manufacturing and infrastructure upgrade trend.
How to Track Global AI Semiconductor Opportunities with Gate Stock Trading
As the AI supply chain continues to expand, investors are broadening their focus from single AI leaders to global tech asset allocation. Gate’s stock trading platform covers major equity markets, allowing investors to explore opportunities across different regions, including U.S. AI chipmakers, Korean memory companies, and global semiconductor supply chain firms.
For AI-themed investing, understanding the structure of the supply chain is more important than simply chasing hot stocks. From GPUs to HBM, from high-speed interconnects to data centers, AI value is being unlocked at multiple stages. Future market opportunities may arise from different segments of the supply chain, rather than being concentrated in a single area.
Conclusion
The recent volatility in South Korea’s stock market and large-scale foreign selling have sparked debate over the future of the AI semiconductor rally. However, capital flows show that foreign investors have not abandoned AI—they are simply restructuring their portfolios.
The reduction in SK Hynix holdings reflects a phase of adjustment for high-valuation AI assets, not the end of the HBM investment thesis. As the AI industry moves into a deeper stage of development, the market’s focus is shifting from individual chips to a comprehensive infrastructure system. In the future, AI competition will not only be about computing power, but also about storage, networking, data centers, and energy capabilities. For investors, understanding the long-term evolution of the AI supply chain is far more important than short-term market swings.
FAQs
Q1: Why have foreign investors recently sold large amounts of Korean stocks?
The main reasons include profit-taking, increased market volatility, and portfolio adjustments related to AI semiconductor holdings.
Q2: Does the reduction in SK Hynix holdings by foreign investors mean they are bearish on HBM?
Not necessarily. SK Hynix remains a key player in the HBM space; the reduction likely reflects short-term capital adjustments.
Q3: Why are foreign investors selling Korean stocks but buying U.S. semiconductor ETFs?
This suggests that capital is seeking to reduce single-stock risk while still participating in global AI semiconductor growth opportunities.
Q4: What is the growth outlook for HBM?
As AI models scale up and data center construction continues, high-performance memory demand still has significant growth potential.
Q5: What will be the next investment focus in AI semiconductors?
The market is likely to focus more on high-speed interconnects, data centers, power infrastructure, and the entire AI supply chain moving forward.
