Summary of Key Points from the Conference Call Industry Overview - The conference call primarily discusses the **AI computing power industry**, focusing on the demand for AI training and inference capabilities, particularly in the context of major tech companies like **OpenAI**, **Google**, **Microsoft**, and **NVIDIA** [1][2][3]. Core Insights and Arguments 1. **Strong Demand for AI Computing Power**: There is a robust demand for AI training computing power, with major tech companies increasing their capital expenditures. OpenAI and others are actively building large-scale clusters, with Musk's cluster reaching **350,000 cards** [1][2]. 2. **Shift in AI Narrative**: The narrative in the AI sector is shifting towards **new skilling**, focusing on model iteration paths and training needs. The relationship between TOKEN volume and computing power demand is not linear; a doubling of TOKEN volume could lead to an exponential increase in computing power needs, potentially by **ten times or more** [1][4]. 3. **Emerging Trends in AI**: The second half of 2025 is expected to see a rise in training demand, with inference demand already increasing. Major cloud service providers are becoming key suppliers as companies like OpenAI and Google collaborate on TPU procurement [1][7][9]. 4. **Market Performance Indicators**: The recent highs in the US stock market reflect changes in the industry, with companies like **Microsoft**, **Broadcom**, **NVIDIA**, and **TSMC** showing strong performance, validating the investment logic in the overseas computing power chain [1][8]. 5. **Investment Focus**: Investors should focus on the changes in the US AI landscape and consider purchasing related assets or companies, as the gap in foundational model capabilities between the US and China may widen [1][6]. Additional Important Insights 1. **Token Volume and Computing Demand**: The increase in TOKEN volume does not necessarily correlate with an increase in computing demand due to the variable of unit TOKEN cost. However, the usage of TOKENs in applications like chatbots and deep research is expected to significantly increase computing demand [23][24]. 2. **Challenges in Tracking AI Demand**: Tracking training demand is more complex than tracking inference demand, requiring an understanding of large model evolution and vendor strategies. Observing the establishment of large clusters can indicate effective training capacity investments [12][13]. 3. **Future AI Trends**: The AI industry is expected to see multiple new trends emerge in the second half of 2025, with a focus on the US market's model and application developments. Investors in the Asia-Pacific region should adjust their strategies accordingly [3][18]. 4. **Key Validation Points**: Important upcoming events include performance forecasts from major domestic companies in July, capital expenditure data from Microsoft, Amazon, and Meta, and NVIDIA's quarterly report, which will help validate training and inference demand [37]. Conclusion The AI computing power industry is experiencing significant changes driven by increased demand for training and inference capabilities. Investors should remain vigilant about market trends, focusing on the US landscape and the implications of TOKEN volume on computing power needs. The upcoming months will be critical for validating these trends and shaping investment strategies.