Macro Analyst: Zhou Junzhi, Jiang Jiaxiu

English summary

1. Under the extremely crowded AI hardware trade, debt financing concerns, the rising competitiveness of Chinese large models, and the breakthrough of domestic equipment have combined to shatter the narrative of AI hardware's 'pricing bottleneck.' Coupled with deleveraging, global AI trades have seen a broad pullback. 2. How has the narrative changed? Since the beginning of this year, hardware excess returns have been built on two pillars: pricing power from supply bottlenecks and downstream's willingness to buy regardless of returns. Challenges to cloud vendors' ROI and financing cash flow, as well as competition from Chinese large models and equipment, have marginally loosened the market's assumptions on these two pillars. 3. How should we understand the ebb of the hardware rally? The laws of technological revolution show that any hardware link that can be standardized will eventually find it difficult to capture most of the industry's profits. Downstream applications with scale advantages and user access will instead enjoy the lion's share of industry profits. 4. How to view the direction of the technology rally? The endgame of the large model business model is a low-margin business with sufficient scale, 'earning moderate gross profit on trillion-dollar revenue' (Altman). ...