平安基金翟森：AI泡沫论的误区与现实

Core Viewpoint - The discussion around whether an "AI bubble" is forming is intensifying in global capital markets, with some arguing that the current situation mirrors the 2000 internet bubble due to rapid CapEx expansion by tech giants and insufficient short-term revenue realization. However, advancements like Google's Gemini 3 suggest that this "bubble theory" is increasingly disconnected from reality, as AI represents a fundamental restructuring of productivity rather than a mere upgrade of information flow [1][2]. Group 1: AI Investment Cycle vs. Internet Investment Cycle - The core logic of the "bubble theory" is that rapid CapEx expansion by companies leads to insufficient revenue realization, making investments unsustainable. This logic applies to the internet model, where growth relies on user behavior and online engagement. In contrast, AI's growth is driven by computational capacity and model capabilities rather than user scale [2][3]. - AI's growth path is more akin to infrastructure investments like electrification and early cloud computing, rather than traffic-driven investments seen in advertising and e-commerce [2]. Group 2: AI Commercialization Logic - Many discussions around AI commercialization are flawed because they still rely on the "user scale—traffic monetization" framework from the past two decades. AI's value is determined by its ability to complete tasks for organizations, not by user engagement metrics [3][4]. - The payment structure for AI differs significantly from the internet. Enterprises are the largest payers, investing in process automation and productivity improvements. The advertising model will shift from content distribution to efficiency metrics as AI capabilities improve [3][4]. - AI's commercialization structure consists of three layers: enterprise automation, advertising efficiency, and consumer value-added services, with the enterprise segment offering significantly greater value than historical internet monetization models [3][4]. Group 3: Current State of AI Commercialization - AI commercialization is still in its early stages, particularly in the consumer sector, and has not yet established large-scale, repeatable product models. The revenue from AI is inherently lagging as organizations must overhaul existing processes and structures to integrate new tools [4][5]. - The mismatch between AI investment and revenue is not a sign of a bubble but a common characteristic of productivity revolutions, where capacity is expanded before large-scale commercialization occurs [5]. Group 4: Model Capability and Future Outlook - The recent release of Google's Gemini 3 marks a significant leap in AI's ability to execute complex workflows, indicating a shift from merely answering questions to executing tasks. This transition suggests that AI will soon directly contribute to productivity [6]. - The true long-term value of AI is driven by productivity improvements rather than short-term revenue. As AI transitions from a focus on generation to execution, it signifies a fundamental restructuring of production methods, with the current commercialization mismatch being a normal delay between technological advancement and organizational adaptation [6].