从狂热到回撤：AI硬件的第一次真实压力测试

Core Viewpoint - The current AI market is more about capital investment than a true productivity revolution, with fears centered on capital withdrawal rather than technological stagnation [1] Group 1: AI Hardware Market Dynamics - The AI market over the past two years has been a revaluation of computing infrastructure rather than a pure AI boom, with companies like Bloom Energy and Micron Technology benefiting as "shovel sellers" [3] - Short-term demand for AI hardware is driven by exponential growth in computing power needs, leading to increased demand for electricity, optical communication, and data center infrastructure [6] - Long-term sustainability of this demand hinges on the ability of AI models to generate stable commercial returns, which is currently lacking as many AI applications remain in low-cost acquisition phases [6][7] Group 2: Market Corrections and Expectations - The rapid rise of AI hardware stocks has led to significant corrections as market expectations shift, particularly regarding capital expenditure and supply-demand dynamics [9][10] - The first correction stems from changing expectations about capital expenditure as the profitability of model companies comes into question, leading to a reassessment of future computing investments [10] - The second correction is due to supply-side adjustments, where previously constrained supply is beginning to meet demand, leading to price elasticity changes [10] - The third correction involves a return to valuation fundamentals, where high growth expectations are recalibrated, causing stock prices to drop even without a fundamental deterioration [11] Group 3: Investment Strategy Shifts - Investors need to transition from a "story-driven" valuation to a "cash flow-driven" approach, recognizing that AI models require longer return cycles and higher capital consumption [13][14] - Investment focus should shift from sector selection to company selection, prioritizing firms with cost advantages and stable cash flows to navigate through market cycles [15] - Understanding the cyclical nature of AI hardware is crucial, with investment strategies needing to adapt based on current market conditions rather than solely future potential [15][16] Group 4: Structural Changes and Future Opportunities - The long-term logic of AI hardware exists but will manifest through complex and volatile patterns rather than linear growth, with structural changes providing clues for navigating cycles [17] - The AI landscape is characterized by a cycle of "explosion—overcapacity—clearing—rebalancing," where companies maintaining healthy balance sheets during downturns will capture greater market share in subsequent cycles [16][17] - The ultimate winners in the AI revolution will be those who create real value rather than merely consuming capital, emphasizing the importance of endurance in the market [19][20]