大摩：市场低估了明年潜在的“AI重大利好”，但存在关键的不确定性

Core Viewpoint - A significant leap in AI capabilities driven by exponential growth in computing power is anticipated by 2026, which may be underestimated by the market [5][6]. Group 1: Computing Power Growth - Major developers of large language models (LLMs) plan to increase their computing power for training cutting-edge models by approximately 10 times by the end of 2025 [5]. - A data center powered by Blackwell GPUs is expected to exceed 5000 exaFLOPs, significantly surpassing the computing power of the U.S. government's "Frontier" supercomputer, which is slightly above 1 exaFLOP [8]. - The report suggests that if the current "scale law" continues, the consequences could be seismic, impacting asset valuations across AI infrastructure and global supply chains [6][8]. Group 2: Scaling Wall Debate - The concept of the "Scaling Wall" indicates that after a certain threshold of computing power investment, improvements in model intelligence and creativity may diminish rapidly, posing a significant uncertainty in AI development [10]. - Recent research indicates that using synthetic data for large-scale training did not show foreseeable performance degradation, suggesting that the risk of hitting the "Scaling Wall" may be lower than expected [11]. Group 3: Asset Valuation Implications - If AI capabilities achieve a nonlinear leap, investors should assess the multifaceted impacts on asset valuations, focusing on four core areas: 1. AI infrastructure stocks, particularly those alleviating data center growth bottlenecks [13]. 2. The U.S.-China supply chain, where intensified AI competition may accelerate decoupling in critical minerals [14]. 3. Stocks of AI adopters with pricing power, which could create an estimated $13 trillion to $16 trillion in market value for the S&P 500 [14]. 4. Long-term appreciation of hard assets that cannot be easily replicated by AI, such as land, energy, and specific infrastructure [15].