留存到底好不好，才证明这些公司有没有后续的发展潜力。 在2025Inclusion·外滩大会上，当被问及如何判断AI时代创业公司是否值得投资，金沙江创投主管合伙 人朱啸虎给出了一个简洁的回答："我们一直只看一个指标，就是留存。" 在他看来，AI时代的创业公司与以往PC互联网、移动互联网时代并没有本质不同，真正能够生存和发 展的企业，仍然需要回归商业本质。他尖锐指出，今年很多被嘲笑的AI公司正是因为缺乏用户留存。 "AI时代很多人都愿意尝鲜，任何新事物出来之后大家一开始觉得好用，第一个月就付了钱，但第二个 月也许就不付钱了，没有留存。"朱啸虎说，如果重新把一个用户召唤回来，在移动时代需要10倍以上 的成本，几乎没有可能性，所以关键还是留存。"留存到底好不好，才证明这些公司有没有后续的发展 潜力。" 至于看好哪类AI应用的机会，朱啸虎观察到一个有趣现象：真正适合商业化的都不是那么"性感"的技 术，而是看上去已经比较无聊的技术。 例如，去年全世界做得最好的AI商业化就是各种各样的会议纪要。无论是美国的Abridge（做医生和病 人之间谈话的会议纪要），还是中国的Plaud（可以磁吸手机的AI录音卡片），这些技术没有 ...

Core Viewpoint - The AI industry continues to be a focal point in global technology and capital markets, with evolving development paths and investment directions [1] Group 1: Investment Opportunities - Investment opportunities in the AI sector by 2025 can be centered around three main logics: continuous overseas computing power investment, breakthroughs in domestic large models, and the imminent explosion of AI applications [4] - Major overseas tech companies are significantly increasing their cash flow investments in AI computing power, with Meta's investment cash flow ratio rising from 25%-30% to 92.7% by mid-2025, indicating intense competition for computing resources [4][5] Group 2: Domestic Market Challenges - The development of the domestic computing power chain has been volatile, with capital expenditures from domestic cloud vendors experiencing fluctuations, leading to a lack of significant advancements in large model development compared to overseas counterparts [8] - The primary challenge for domestic computing power lies in the semiconductor sector, where high-end training and inference chips are limited, constraining the willingness and ability of domestic cloud vendors to expand computing power [10] Group 3: Semiconductor and AI Applications - The semiconductor industry's breakthrough, including design, manufacturing, and packaging, is crucial for the rise of the domestic AI ecosystem, with a systemic opportunity expected to emerge by the end of this year or next year [10] - AI applications have yet to produce a truly groundbreaking product, primarily due to a mismatch between technological capabilities, product forms, and user needs [11] Group 4: Key Application Areas - Notable AI application areas include AI smartphones, AR glasses, autonomous driving, and humanoid robots, with the fourth quarter of 2025 expected to be a critical validation point for multiple AI applications [12] - The financial sector shows significant demand for AI talent, as AI technologies can enhance efficiency in areas like smart investment advisory, risk control, and customer service, particularly in a favorable market liquidity environment [12] Group 5: Long-term Outlook - Overall, the three trends in the AI industry are moving positively: continuous growth in overseas computing power investment, acceleration of domestic semiconductor self-sufficiency, and AI applications nearing an explosive phase [13] - Investors are encouraged to focus on different segments based on these trends, such as overseas computing chains and breakthroughs in domestic high-end chips, while closely monitoring product validation in the application sector [13]