2026-02-02 02:15Summary of Conference Call Records Industry Overview - The focus is on the overseas AI industry, particularly the demand for computing power related to AI inference and training, which has recently strengthened. [1][2] Core Insights and Arguments - Recent price increases by Amazon Cloud and Google Cloud indicate a rising demand for AI computing power, with TSMC revising its capital expenditure (Capex) upwards. [1] - Despite limited visibility for large-scale commercialization of AI applications, the demand for computing power is expected to rise further in the next 3-6 months, alleviating concerns about a "computing power bubble." [1] - The emergence of new generation agents like MoltBot is significantly increasing the consumption of inference computing power, enhancing capabilities for complex tasks. [1] - Anthropic is expanding its product offerings with Claude Code and Agent Skills, broadening the application scenarios for agents. [1] Training Demand Insights - The token call volume has been rapidly increasing since early January 2026, indicating a strong upward trend in inference demand, with Anthropic and its cloud service providers likely to be key beneficiaries. [2] - A new wave of models is expected to be released in Q1 2026, with advancements in language models like Grok-5 and GPT-6, as well as rapid iterations in multimodal models such as Veo-4, which will place higher demands on training computing power. [2][3] Financial Catalysts - The upcoming earnings reports from major companies like Meta, Google, and Amazon are anticipated to validate the trends in computing power demand and Capex. [3] - Nvidia's earnings report and the GTC conference in March are expected to further reinforce the investment outlook for computing power throughout the year. [3] Investment Strategy - The demand for computing power is expected to maintain an upward trajectory, leading to a potential recovery in sector sentiment. [3] - Three key investment opportunities are identified: 1. Cloud Providers: Companies like Amazon and Google are expected to benefit from increased inference demand driven by agents. [3] 2. Overseas Computing Power Chain: Companies such as Nvidia, Zhaoyi Innovation, and others are recommended due to their potential for revaluation amid model iteration. [3] 3. Model Companies: Firms like Meta, Google, Alibaba, Tencent, and Minimax are highlighted for potential valuation reappraisal due to exceeding expectations in model capabilities. [3]