CITIC Securities: Still optimistic about US stocks, the main technology line currently favors the middle and downstream links of overseas computing power chains

The Zhitong Finance App learned that CITIC Securities released a research report saying that in the process of popularizing AI, end users do not always choose the model that consumes the most computational power and has the best performance, and often face financial constraints on token budgets. The strong performance of computing power hardware in the second quarter of this year was mainly driven by price increases rather than volume. As the market focus changed from scarcity to sustainability, overseas markets represented by Korean stocks recently faced adjustments. Computing power is not excessive, and the evolution of bargaining power is becoming more and more critical. It is noted that the AI adoption rate of American companies is increasing, and the overall investment intensity of the enterprise sector in technology is in line with long-term trends. I believe that the market competition mechanism will push more enterprises to use AI to improve work efficiency, which will bring about demand for new computing power services, AI models, and applications. I am still optimistic about the main technology line of US stocks. Currently, I prefer the middle and downstream links in the overseas computing power chain, and I am optimistic about the attractiveness of US stocks compared to the Japanese and South Korean stock markets.

CITIC Securities's main views are as follows:

Recently, doubts about the sustainability of AI narratives have intensified in overseas markets.

US stocks emerged from a structured market driven by technology in the first half of this year. The stock prices of companies in various positions in the AI industry chain began to diverge in June. Computing power hardware was the star sector that attracted attention in the global market in the second quarter. However, since late June, there has been endless news about market fluctuations in the technology sector, and fluctuations in overseas markets, represented by the Korean stock market, have become drastic, and the consensus that global capital “groups” semiconductors has also begun to loosen.

The strong performance of overseas computing power hardware is mainly driven by price increases rather than volume, which is typical cyclical logic.

The strong momentum of technology hardware in the second quarter was mainly driven by rising product prices rather than a sharp rise in deliveries. South Korea's semiconductor exports even saw a 25% year-on-year decline in May of this year. This shows that the trading basis of the recent market comes from high premiums due to physical scarcity. Compared with previous storage cycles and the trading logic of traditional commodities, they are all essentially phased mismatches between supply and demand. However, the upward period of this cycle may be longer and the space may be higher. The market's expectations for overseas semiconductor manufacturers are already optimistic, and the evolution of the latter's bargaining power is worth watching.

As the market focus shifts from scarcity to sustainability, the financial accounts of AI investments become more important.

The core variable that determines the credibility of AI narratives is the prospects for commercial application of terminals. Admittedly, AI models tend to perform better when they consume more computing power, but token budgets should be a more critical factor in the AI investment behavior of terminal companies. It is difficult for ordinary users to ignore their financial costs just to chase high performance. This will eventually influence upstream bargaining power through cloud factories' willingness to spend capital. It is noted that the AI adoption rate of US companies is increasing, and that the overall investment intensity of the enterprise sector in technology is in line with long-term trends and is not financially overdrawn to future needs. I believe that the market competition mechanism will push more enterprises to use AI to improve work efficiency, which will bring new computing power services, AI models and application requirements. Compared to the high expectations of a sharp rise in upstream volume and price in overseas computing power chains, it endorses the deductive logic of the middle and lower reaches of the “Jevens Paradox”.

After the ultimate market, find a more flexible trading direction.

Currently, the flexibility of the traditional pro-cyclical sector of US stocks is inevitably hampered by high interest rates, while technology investment has always been less sensitive to changes in interest rates, and is still more optimistic about opportunities in the main technology sector. Against the backdrop that overseas technology hardware stock prices have mostly risen significantly, and non-hardware performance is slightly poor, the latter seems more likely to bring fundamental surprises. Currently, they prefer the middle and downstream links of the computing power chain. This logic also applies to the trade-off between US stocks and the Japanese and South Korean stock markets. If the upstream supply and demand gap in the computing power chain tends to ease, and if the decline in token prices and market competition mechanisms also encourage end users to increase the usage of AI services, then US stocks may once again become more attractive than the Japanese and South Korean stock markets. After many tests and contacts, the US and Iran should have gradually become familiar with each other's intentions and bottom line. The most intense period of conflict is over. In anticipation of moderate US economic growth and insufficient liquidity, they are still optimistic about the allocation value of US stocks compared to US debt.

Risk factors:

Changes in global market liquidity or sentiment exceed expectations; technological progress, corporate performance, policy changes, or unexpected events affect expectations; overseas economic momentum falls short of expectations or inflation exceeds expectations.