瑞银证券熊玮：现阶段中国出现“AI泡沫”可能性不大

Core Viewpoint - UBS Securities analyst Xiong Wei believes that by 2026, the application cases of AI in China will become more abundant and the monetization pace will accelerate, indicating a low likelihood of an "AI bubble" in the current stage [1]. Group 1: AI Development in China - By 2025, investor interest and recognition of AI development in China are expected to increase due to advancements in large model technology, computing infrastructure, and the widespread adoption by enterprises and consumers [1]. - In 2026, domestic large model capabilities are anticipated to rapidly iterate, catching up with their American counterparts, while both China and the U.S. will follow similar monetization paths in AI applications, particularly in cloud services and advertising [1]. - The localization of computing power is expected to continue, with breakthroughs in chip performance, leading to supernodes taking a larger share of inference and training workloads [1]. Group 2: Factors Mitigating AI Bubble Risks - The limited cycle of financing is a key reason for the low likelihood of an "AI bubble," as leading AI model developers are funded by internal cash flows from parent companies rather than third-party financing [1]. - Major Chinese internet companies are adopting a pragmatic approach to AI investments, focusing on return on investment and operational efficiency rather than excessive initial spending [1]. - The stability of the Internet Data Center (IDC) deployment rate and regulatory control over new supply, including strict management of electricity quotas by the National Development and Reform Commission, help prevent overbuilding [1]. Group 3: AI Disruption in Vertical Industries - The discussion around AI's potential to disrupt vertical industries is increasing among investors, but the pace of disruption in China may remain slow due to a more fragmented landscape of AI chatbots compared to the dominance of ChatGPT and Gemini in the U.S. [2]. - Vertical industry companies are actively integrating AI and intelligent agents into their applications, but the high entry barriers in sectors like online travel agencies (OTA) and online music content and user operations make it difficult for AI to easily replace existing capabilities [2].