EB SECURITIES: Demand for AI computing power in the US stock market is heating up, and the bullish sentiment driven by AI is once again active.

EB SECURITIES released a research report stating that since 2Q25, the US stock AI sector has been continuously strong, with AI application-focused companies seeing valuation recovery and enjoying a higher premium compared to the software industry. From April 9, 2025, to July 28, 2025, tariff concerns eased, cloud vendors in the US signaled strong demand for AI computing power, and the market value of AI computing power-focused companies rose by as much as 102%. EB SECURITIES selected 66 key AI companies from 25M4 to 25M7. The valuation gap between key AI application companies and S&P North American technology software companies has returned to historical highs, indicating that market optimism driven by AI is once again active. EB SECURITIES' specific analysis is as follows: What industrial trends does the leading power reflect? 1) Behind the recent AI computing power market, there is long-term industrial trend support, demonstrating the multidimensional market demand for global AI applications. The potential space for long-tail computing power demand far exceeds market expectations; Cost-effective lightweight AI models are gradually favored by the market; The space for sovereign AI computing power markets is vast. 2) The fundamentals of the software industry are stable, trending upwards but with limited marginal improvement. Revenue: The proportion of companies exceeding expectations is at a historical high, and the median level of exceeding expectations is relatively stable, indicating steady performance, but AI has not brought a clear increase in revenue exceeding expectations. EPS: The proportion of companies exceeding expectations is on the rise, and the median level of exceeding expectations is decreasing. The software industry has moved away from the profit volume cycle, moving towards the AI investment and expansion cycle. Analyze the core growth engine of AI computing power from multiple dimensions of technology, cost, and demand. 1) Large Models: Cost reduction drives rapid growth of API calls, validating the Jevons paradox. Since 2025, the rapid growth of API calls for large models is mainly driven by the demand for lightweight large models, while the demand for high-performance large models is relatively stable, usually for iteratio...