The capital that had been betting on South Korea's AI narrative for half a year may now be executing a large-scale retreat.
On July 9th, the Korea Composite Stock Price Index (Kospi) fell 5.4% in a single day, marking a cumulative decline of approximately 20% from its record high last month and officially entering a technical bear market. This turn of events is particularly dramatic. Earlier this year, the Kospi was among the world's best-performing major indices, having surged as much as 116%. That gain has now moderated to about 72%.
Amid the South Korean market plunge, Goldman Sachs' thematic research team released a report, advising clients to shift their positioning from Korean AI trades to the "China AI value chain." Yesterday, capital accelerated its exit from South Korean chip stocks, flowing into Hong Kong's tech sector. The Hang Seng China Enterprises Index rose as much as 4.5% intraday, with Alibaba's Hong Kong shares surging over 13%.
South Korean Stocks: From Top Performer to Bear Market
The Kospi index fell 5.4% on Wednesday. Leading the decline were the very stocks that had previously driven the index's surge—SK Hynix Inc. fell 5.7% and Samsung Electronics Co., Ltd. dropped 6.3%.
Ironically, Samsung had just reported quarterly profits soaring 19-fold year-over-year this week, yet its stock still tumbled. This indicates market concerns are no longer about current profitability, but about the future: Can the AI capital expenditure boom be sustained?
Ian Samson, a portfolio manager at Fidelity International, pinpointed the issue: "A lot of the volatility is coming from uncertainty around the fundamentals. We do see AI-driven semiconductor demand as real and massive — but it's essentially being supported by about $1 trillion in capital expenditure controlled by just a handful of large tech companies." Once this spending slows, the downside risk will materialize quickly.
Another structural vulnerability in the South Korean market is the heavy use of leveraged ETFs by retail investors to concentrate on chip stocks, which can amplify declines when the trend reverses, as reported by Bloomberg. Foreign investors have also been persistent sellers—global funds have offloaded over $100 billion worth of South Korean local stocks so far this year.
Additionally, media reports suggest the rise of domestic Chinese chips. There have also been reports that Apple is lobbying the U.S. government for permission to purchase memory chips from ChangXin Memory Technologies (CXMT).
Where is the Money Flowing?
The flow of capital tells the story.
The Hang Seng China Enterprises Index jumped 4.5% on Wednesday, its largest single-day gain since February 2025. Alibaba's Hong Kong-listed shares surged over 13%, while Tencent Holdings Ltd rose more than 4%. Month-to-date, the Hang Seng series of indices are the best-performing benchmarks in Asia, while the Kospi is the worst.
Gerald Gan, CEO of Reed Capital, stated: "The performance divergence between China and other global markets is particularly stark, creating a compelling value opportunity for Chinese equities. Major Chinese tech companies are exactly the targets we are accumulating positions in."
He also noted that the AI-driven rallies in South Korea and Taiwan "may be showing signs of fatigue," and investor awareness of portfolio concentration risk is increasing. "It is reasonable to rebalance at this juncture."
Goldman Sachs Weighs In: Recommends the "China AI Value Chain"
This capital rotation has received Goldman Sachs' endorsement.
The bank's research team recently published a report with the direct title: "Trade Idea: Long China AI Value Chain." Analyst Louis Miller, the report's author, wrote: "China AI is now formally on our radar."
Goldman Sachs recommends clients buy its proprietary "GS China AI Value Chain" basket product, which covers the entire industrial chain from power and semiconductors to AI infrastructure, AI models, and AI applications.
The analysts present three core rationales:
First, Chinese AI valuations are severely depressed. Since late 2022, global AI-related stocks have collectively created $34 trillion in market value, with China's share being minimal. The current market value of China's AI-related sector is approximately $4 trillion—analysts believe this figure is "significantly low" relative to China's actual position in the global AI industry.
Supporting data: China accounts for 10% of global AI-related market value and 16% of AI-related revenue, but global mutual fund managers' allocation to Chinese tech stood at just 1.2% as of January 2026.
The bank's research also estimates that the potential economic benefits from AI-driven efficiency gains and new profit creation could be 50% to 100% higher than the expectations implied by current AI stock prices.
Second, China's structural advantages are undervalued by the market. The analysts believe China possesses competitive advantages in the infrastructure, power, and semiconductor segments of the AI supply chain, but these are not yet fully priced in.
Third, Chinese AI stocks are already outperforming other Chinese assets, but still show a significant gap compared to U.S. AI stocks. This implies room for catch-up remains.
Supporting Rationale: Multiple Catalysts in Concert
The bank argues that this round of Chinese AI performance is structural, not a fleeting rebound, and lists several specific supports:
Soaring chip exports: China's chip sales in May surged 111% year-over-year, with overall exports growing 19.4%, the strongest in nearly three months, primarily driven by AI hardware demand.
Index reconfiguration: Major Chinese exchanges are adjusting benchmark indices to include more domestic AI and semiconductor companies, which will direct passive funds toward strategic tech sectors.
IPO acceleration: ChangXin Memory Technologies has officially received approval to list on the Shanghai Stock Exchange's STAR Market, becoming one of the largest A-share IPOs this year.
The bank specifically emphasizes that this is not a KWEB (China Internet ETF) trade, but a structural opportunity covering the entire AI industrial chain.