CITIC SEC has released a research report stating that during the process of AI adoption, end-users do not always choose the most computationally intensive and highest-performing models, often facing financial constraints related to token budgets. The strong performance of computing power hardware in the second quarter of this year was primarily driven by price increases rather than volume growth. As market focus shifts from scarcity to sustainability, overseas markets, represented by South Korean stocks, have recently undergone adjustments. Computing power is not in surplus, and the evolution of pricing power is becoming increasingly critical.
The report notes that AI adoption rates among US companies are rising, and the overall intensity of technology investment in the corporate sector aligns with long-term trends. It is believed that market competition mechanisms will drive more companies to use AI to enhance work efficiency, generating new demand for computing power services, AI models, and applications. The firm remains optimistic about the technology theme in US stocks, currently showing a preference for the mid-to-downstream segments of the overseas computing power chain, and views US equities as more attractive compared to Japanese and South Korean markets.
Key Points from CITIC SEC's Report
Recent concerns about the sustainability of the AI narrative have intensified in overseas markets. US stocks experienced a structurally driven rally in the first half of the year, led by the technology sector. However, stock prices of companies across the AI industry chain began to diverge in June, with computing power hardware being the star segment globally in Q2. Since late June, numerous news events in the tech sector have caused market volatility, leading to sharp fluctuations in overseas markets, notably South Korean equities. The global consensus of capital concentrating in semiconductors also appears to be loosening.
The robust performance of overseas computing power hardware has been primarily driven by price hikes rather than shipment volume growth, following a typical cyclical logic. The strong momentum in tech hardware during Q2 was mainly fueled by rising product prices, not a surge in delivery volumes. South Korea's semiconductor export volume even saw a year-on-year decline of approximately 25% in May this year. This indicates that recent market trading has been based on high premiums stemming from physical scarcity, not fundamentally different from the trading logic of past memory cycles or traditional commodities. Essentially, it involves trading on temporary supply-demand mismatches, though this cycle's upswing may last longer and reach higher levels.
Market expectations for overseas semiconductor manufacturers are already optimistic, and the evolution of their pricing power warrants attention. As market focus shifts from scarcity to sustainability, the economic viability of AI investments becomes more important. The core variable determining the credibility of the AI narrative lies in the prospects for end-user commercial applications. While AI models often perform better with greater computational consumption, token budgets are likely the more critical factor influencing end-enterprise AI investment decisions. Ordinary users are unlikely to chase high performance while ignoring its financial cost, which will ultimately affect upstream pricing power through cloud providers' capital expenditure willingness.
Observing that AI adoption rates among US companies are increasing, and overall corporate sector tech investment intensity aligns with long-term trends without financially overextending future demand, the report believes market competition will push more firms to use AI for efficiency gains. This will create new demand for computing power services, AI models, and applications.
Compared to the high expectations for simultaneous volume and price growth in the upstream overseas computing power chain, the report finds the "Jevons paradox" deduction logic in the mid-to-downstream segments more convincing.
Identifying More Flexible Trading Directions After Extreme Market Conditions
Currently, the flexibility of traditional pro-cyclical sectors in US stocks is inevitably constrained by high interest rates. Since technology investment is typically less sensitive to interest rate changes, the firm remains more optimistic about opportunities within the tech theme. Given that most overseas tech hardware stocks have risen significantly while non-hardware performance has been relatively weaker, the latter seems more likely to deliver positive fundamental surprises. The current preference is for the mid-to-downstream segments of the computing power chain.
This logic also applies to the trade-off between US stocks and Japanese/South Korean equities. If the supply-demand gap in the upstream computing power chain eases, token price declines and market competition encourage end-users to increase AI service usage, then US stocks may regain better attractiveness compared to Japanese and South Korean markets.
After multiple rounds of testing and contact, the US and Iran should have gradually familiarized themselves with each other's intentions and bottom lines. The peak of conflict has likely passed. Against expectations of moderate US economic growth and not overly abundant liquidity, the report maintains a positive view on the allocation value of US stocks relative to US Treasuries.
Risk Factors
Global market liquidity or sentiment changes exceeding expectations; technological progress, corporate earnings, policy changes, or unexpected events having a greater-than-anticipated impact; overseas economic momentum falling short of expectations or inflation exceeding forecasts.
Comments