This article is written based on public information and is only for the purpose of information exchange and does not constitute any investment advice.

1. The underlying belief of current AI is the Scaling Law, but this law is not a theorem but a statistical rule. Whether it can continue in the future depends more on the underlying beliefs. At present, large-scale model investors are in the trial-and-error stage, and downstream enterprises represented by North American CSPs continue to pay for this belief.

However, it is difficult to predict the final effect of the model, so the market is mainly concerned about the demand in 2025 and beyond. The recent large fluctuations in US technology stocks are closely related to this concern.

2. There may be two reasons for the delay in the release of OpenAI's next-generation GPT-5 product. One is that the number of parameters in the model is too large, which significantly increases the difficulty of training. It is understood that a more significant constraint is data. The amount of existing data is insufficient, and the quality decreases as the data volume expands, interfering with the efficiency and accuracy of model training.

3. Due to policy support, the low-altitude economy has attracted increasing attention, and its future industry potential should not be underestimated. The market space may reach trillions of yuan.

Looking ahead, the development of the industry may be divided into two stages: In the next five years, it will focus on markets with high paying capabilities, such as air tours, flights, and medical transfers, which are "pilot" projects. From 2030 to 2035, it is expected to target mass consumption, such as air taxis and air buses, and we may also see an explosion of eVTOL entering the private market, a to-C scenario.

4. The competition in the mobile SoC market has become more intense, and the competition in AI terminals will only exacerbate this trend.

Facing sluggish growth, Qualcomm has begun to eye MediaTek's market share, sinking into low-end markets such as thousand-yuan phones, increasing rebate rates, and offering significant discounts on older chip models. In addition, in AI terminals, Apple has secured a place, Qualcomm has chosen to collaborate with more vendors in mobile phones and PCs, while MediaTek may partner with NVIDIA.

5. Domestic consumption of sports products is becoming more rational, and the pursuit of value for money is becoming a trend.

Since the beginning of this year, the online growth rate of sports products has been in single digits. Amid declining growth rates, professional, new, and low-priced products have performed better structurally. For example, Anta's affordable and celebrity basketball shoes have sold well, and its sports casual shoes, except for some luxury replacements from FILA, have also performed outstandingly. In addition, Anta has benefited more from the exposure of the Olympics, while Li-Ning has a relatively small proportion of low-end products, and its high-end lines have been significantly impacted.

6. Aier Eye Hospital will face greater challenges as the number of companies in the industry grows, intensifying competition and leading to a decline in profitability. The number of eye hospitals is expected to increase significantly in the future, and some private eye hospitals are also planning to go public. The intensification of industry competition will further reduce hospital gross margins.