华尔街见闻·2026-02-16 11:18Core Insights - The core viewpoint of the article is that the expansion of AI capabilities is significantly constrained by TSMC's conservative capacity expansion strategy, which may lead to a major chip shortage by 2029 if not addressed [2][4]. Group 1: TSMC's Capacity and AI Expansion - TSMC, as a monopolistic player, is cautious in expanding its production capacity due to the high risks associated with wafer fabrication, preferring to avoid the potential for overcapacity and its associated depreciation costs [2][3]. - This conservative approach results in a misalignment of risks, transferring the burden of insufficient capacity to major tech companies like NVIDIA and Apple, which face the risk of losing future revenues due to inadequate computing power [3]. Group 2: Future Predictions - A significant prediction made is that a large-scale chip shortage is expected around 2029, as current capital expenditure growth (e.g., TSMC's increase from $40 billion to $60 billion) is insufficient to meet the exponential demand for computing power driven by AI advancements [4]. Group 3: Recommendations for Tech Giants - Tech giants are urged to support companies like Intel or Samsung, or to take on factory construction risks through prepayments, driven by economic motives rather than geopolitical considerations, to avoid being trapped in a capacity bottleneck [5]. Group 4: Monetization of AI Applications - The article emphasizes that advertising is the most effective monetization method for AI applications, particularly for companies like OpenAI that have significant traffic but lack a solid business model [6]. - Thompson counters the argument that advertising negatively impacts AI answer quality, asserting that a comprehensive understanding of users is essential for effective advertising [10]. Group 5: Analysis of Major Tech Companies - Meta is highlighted as having the strongest execution capabilities, with its advertising model being undervalued despite concerns over capital expenditures [12]. - Google is described as chaotic yet resilient, likened to a slime mold that, while appearing disorganized, possesses great adaptability [13]. - Amazon's strategy in the AI era raises concerns, as its focus on low-cost alternatives may hinder competitiveness in a rapidly evolving market [14]. - Apple is criticized for being a poor platform manager despite its hardware strengths, indicating a need for improvement in software and service platforms [16]. Group 6: Future of SaaS and Value of "Live" Experiences - The article suggests that if AI leads to a reduction in workforce numbers, the SaaS business model based on "per seat" pricing will face growth limitations [18]. - In a world flooded with AI-generated content, the value of "live" experiences, such as shared events and face-to-face interactions, will become increasingly significant [19].