Core Viewpoint - The investment landscape for 2026 is characterized by high stock valuations, driven by significant investments from hyperscalers in data centers and cloud infrastructure, drawing parallels to the 1999-2000 internet bubble [1][2] Group 1: Stock Market Valuation - Current stock valuations are high compared to historical levels but have not reached extreme levels [2] - Some investors are considering reducing risk exposure, particularly those managing defined benefit pension plans, while most investors remain invested due to the need for returns that outpace inflation [2][3] Group 2: Investment Strategies - Passive investment in stock indices is a common choice, but it carries risks due to concentration in a few tech companies, which can lead to significant idiosyncratic stock risk [3] - Active management is suggested as a way to navigate risks, with a focus on calculated risk decisions based on various factors supporting the stock market [3][4] Group 3: Economic Outlook - The risk of a recession in the U.S. is low, with a stable labor market and healthy balance sheets for private enterprises, suggesting positive returns from the stock market [4] - The company is closely monitoring large-cap companies for investment returns, particularly those transitioning from free cash flow generators to significant capital spenders [4] Group 4: Diversification Opportunities - The company is seeking diversification opportunities beyond AI, noting the benefits of regional diversification and strong performance of value investments outside the U.S. [5] - Emerging market bonds are providing better yields compared to developed market bonds, and alternative investments like insurance-linked securities and infrastructure debt may offer additional yield opportunities [5]

尽管人工智能（AI）占据了新闻头条，但我们认为其颠覆性的增长潜力和诱人的投资回报仍被低 估。与此同时，投资者面临着"重蹈覆辙"的风险，即期望通过广泛的指数投资来捕捉投资机会，却难以 区分颠覆者与被颠覆者。 二是技术提升迅速，成本随之降低。自ChatGPT发布以来，大语言模型的进化速度之快令人惊叹。 关键在于，推理模型能够生成自己的训练数据，模型处理的每个思考过程都可以反馈回模型中，从而在 解决问题时实现实时学习。 此外，由于半导体技术的进步，技术成本正在迅速下降，为生成式AI带来了更多盈利应用场景。 就在几年前，推理模型还因其计算密集的特性，成本高得令人望而却步。 三是AI扩展性远超其他技术突破。回顾历史技术周期，无论是通过计算机实现劳动力自动化，还 是知识型工作赋能（涉及思考、分析或创造的任务），都需要新设备和流程。自动化流水线已经存在了 一个多世纪，但发展缓慢，因为在现实世界降本需要时间。计算机赋能的知识型工作也需要设备和技 能，例如，学习打字或使用Excel，虽然这一过程进展很快，但也花了数十年时间。 而如今，我们看到AI正以前所未有的速度增强人们的工作能力。随着我们迈向由智能体主导的世 界（首批应用案 ...