CITIC Securities: Hyperscaler CDS surges + Korean stocks suspended again, Hong Kong stocks continue to attract Asia-Pacific rotation funds

According to the Zhitong Finance APP, CITIC Securities released a research report stating that the overnight U.S. stock market's AI hardware sector experienced a significant correction, primarily due to the market's repricing of Hyperscaler capital expenditure expansion, financing leverage levels, and credit risk expectations, rather than a deterioration in the fundamentals of AI demand. The circular financing model dominated by NVIDIA has become the focus of market attention, but currently, AI chip rental prices remain stable, and industrial demand still shows resilience, with financing capacity not being the main constraint on industry development in the short term. The adjustment in the South Korean stock market is more attributed to Changxin Technology's listing, which has broadened the financing channels for capacity expansion, as well as changes in market expectations brought about by the commercialization progress of domestic DUV equipment, with investors pricing in the potential impact of domestic supply expansion on the industry landscape in advance. In contrast, Hong Kong stocks, benefiting from valuation advantages, improved funding structures, and short covering space, are becoming one of the main directions for global funds to shift from highly crowded trades to undervalued assets.

CITIC Securities' main viewpoints are as follows:

Event: On July 27th, Eastern Time, the U.S. stock market's artificial intelligence hardware sector saw another significant correction, with the Philadelphia Semiconductor Index falling by 2.2%, and SanDisk dropping by 11.0%, while SK Hynix fell by 7.5%. Today, panic selling spread to some Asian markets, with the South Korean Composite Stock Price Index plunging more than 8% at one point, triggering the market's circuit breaker mechanism, and closing down 10.8% for the day; the Nikkei 225 Index also fell by 4.0%.

The disclosure of Google's financial report constituted a key catalyst for the upward revision of this round of capital expenditure expectations, while also laying the groundwork for the subsequent widening of CDS spreads for tech giants.

Before the disclosure of Google's financial report, Bloomberg's consensus forecast at the end of June showed that the total capital expenditure of the four major Hyperscalers was expected to reach $718.3 billion by the end of 2026, and $899.8 billion by the end of 2027. After the release of Google's Q2 financial report, the company raised its full-year capital expenditure guidance for 2026 and signaled that it would further increase investment in AI infrastructure in 2027, leading to a simultaneous upward revision of capital expenditure expectations for all Hyperscalers. As of July 24th, market expectations for the total capital expenditure of Hyperscalers in 2026 and 2027 were adjusted to $731.9 billion and $950.2 billion, respectively, an increase of 1.9% and 5.6% compared to the forecasts at the end of June. This round of upward revision indicates that the market's previous judgment on the peak of AI infrastructure investment was still conservative, with actual expenditure intensity and duration exceeding expectations, thereby intensifying market concerns about the future financing needs, leverage levels, cash flow pressures, and commercialization of tech giants.

The core logic behind the soaring CDS of North American tech giants stems from market concerns about NVIDIA's circular financing model.

NVIDIA helps downstream customers obtain funding by providing large-scale guarantees and financing support, allowing customers to further procure its chips, forming a closed-loop structure where cash flow and credit risk are highly correlated. Coupled with the $500 billion collaboration between Hynix and NVIDIA, as well as NVIDIA's ongoing discussions with OpenAI regarding financing guarantee arrangements of up to $250 billion, the market is concerned that the leverage levels in the financing chain continue to rise, with deep binding of credit relationships among all parties. If any link in this chain encounters risk, it could lead to a chain reaction of risk transmission through the credit chain However, from the demand side, the rental prices of AI chips such as H100 and B200 remain relatively stable, indicating that there are no signs of weakening demand for AI computing power itself. Currently, the risks of cyclical financing are more reflected in market expectations rather than having manifested as actual deterioration in fundamentals. At the same time, looking at the financing channels of Hyperscalers, the diversified financing methods such as equity financing, public bonds, private bonds, and potential government support indicate that the sources of capital expenditure for U.S. Hyperscalers still possess strong sustainability, and short-term financing capability is not a core factor restricting the development of the AI industry.

Korean stocks suffer from the dual impact of a reversal in the storage pattern and the loosening of "oligopoly monopoly" premium.

Today, Korean stocks plummeted, with the core contradiction being a systematic reassessment of expectations for the storage industry chain. On one hand, Changxin Technology has officially landed on the A-share market, opening up financing channels for subsequent capacity expansion, thus creating a medium- to long-term imagination space for domestic storage capacity expansion and import substitution; on the other hand, The Information reported that domestic immersion DUV lithography equipment has entered the commercialization stage. If verified, the equipment constraints for domestic storage manufacturers will significantly loosen, raising the ceiling for capacity expansion and fundamentally shaking the logic of tight supply constraints in storage. Against this backdrop, investors are choosing to price in the anticipated increase in supply of Chinese storage chips in advance to hedge against the tail risks of Korean giants. On a deeper level, the "hegemonic premium" previously priced into the technology supply chains of the U.S., Korea, and Europe—based on oligopoly monopoly and technological blockade—also faces systemic correction in the context of continuous breakthroughs in the domestic supply chain.

Momentum trading switches, valuation troughs and micro improvements drive capital back to Hong Kong stocks.

In contrast to the sharp decline in Korean stocks, Hong Kong stocks are becoming the main beneficiaries of the current global capital rebalancing. In the first half of the year, the crowded trades in the AI hardware chain continued to clear, and the profits from the technology sector urgently need to find new allocation directions, while Hong Kong stocks happen to have significant absorption advantages. On one hand, after a period of deep adjustment, the market's pessimistic sentiment has been fully priced in. Even though there has been some rebound since July, the dynamic price-to-earnings ratio of major Hong Kong stock indices remains below historical averages, and the valuation cost-effectiveness is still prominent; on the other hand, the micro structure of Hong Kong stocks is also improving simultaneously—high shareholder returns combined with significant digestion of concerns over lock-up selling pressure are constructing a positive cycle in market liquidity. After the trend of RMB appreciation stabilizes, the willingness of southbound capital to allocate has also begun to recover; at the same time, short-selling transactions and open short positions remain high, making it easier to cover shorts than to add new ones, and short covering itself constitutes incremental buying. Overall, Hong Kong stocks have the conditions to continuously absorb the global momentum trading switch, and the trend of capital inflow is expected to continue.

Risk factors: Global geopolitical conflicts escalate again; deterioration of U.S.-China relations; global central banks tightening monetary policy more than expected; AI commercialization progress falling short of expectations; credit spreads of U.S. tech giants widening more than expected