这是孙正义首次就11月份软银意外披露的清仓英伟达之举置评，他同时驳斥了关于人工智能投资泡沫的 说法。 孙正义表示，"我一股也不想卖的，只是更需要资金来投资OpenAI和其他项目，我是哭着卖出英伟达股 票的。" 他周一在东京的一个论坛上表示，软银那么做只是因为需要募集资本来为数据中心建设等项目供资。 软银集团创始人孙正义表示，若公司拥有无限的资金来支持其在人工智能领域的下一轮投资，包括对 OpenAI的大额押注，那么他本不会卖出英伟达的股份。 孙正义还表示，谈论人工智能投资泡沫的人"不够聪明"。他指出，如果人工智能长期来看能创造全球 GDP的10%，那么即使累计投入数万亿美元，也完全值得。"泡沫在哪里？"他问道。 ...

软银集团创始人孙正义表示，若公司拥有无限的资金来支持其在人工智能领域的下一轮投资，包括对OpenAI的大额押注，那么他本不会卖出英伟达 的股份。 孙正义表示， "我一股也不想卖的，只是更需要资金来投资OpenAI和其他项目，我是哭着卖出英伟达股票的。" 孙正义还表示，谈论人工智能投资泡沫的人"不够聪明"。他指出，如果人工智能长期来看能创造全球GDP的10%，那么即使累计投入数万亿美元，也 完全值得。"泡沫在哪里？"他问道。 这是孙正义首次就11月份软银意外披露的清仓英伟达之举置评，他同时驳斥了关于人工智能投资泡沫的说法。 他周一在东京的一个论坛上表示，软银那么做只是因为需要募集资本来为数据中心建设等项目供资。 ...

Group 1 - The AI sector is experiencing a significant surge, with nearly 20 stocks hitting the upper limit and many large-cap tech stocks like Industrial Fulian, ZTE, and SMIC showing strong performance [1][2] - OpenAI's new audio and video generation model, Sora 2, has been released, demonstrating improved accuracy and realism, and has quickly become the top free app in the Apple Store [2] - CoreWeave's CEO argues that the infrastructure investments by major tech companies are driven by real demand, countering the notion of a "circular" flow of funds among companies [6] Group 2 - The Bank of England and IMF have raised alarms about potential AI investment bubbles, comparing current tech stock valuations to the peak of the dot-com bubble [4][5] - Oxford Economics has identified signs of a bubble in the AI sector, noting that tech stocks account for nearly 40% of the S&P 500, with valuations significantly exceeding reasonable values [4] - Some industry experts, including Jeff Bezos, view the current AI trend as an "industrial bubble" rather than a financial one, suggesting that it will ultimately lead to beneficial innovations [6]

Core Insights - Meta has paused hiring in its AI department after an aggressive recruitment strategy aimed at competing with OpenAI and Google, leading to significant employee turnover [1][2][3] - The recruitment freeze is described as a strategic adjustment to build a more stable structure following a massive talent acquisition effort [3][31] - The internal restructuring involves splitting the AI department into four specialized teams, each with distinct missions, to enhance efficiency and collaboration [32][33] Recruitment Strategy - In June 2025, Mark Zuckerberg invested nearly $15 billion to acquire a 49% non-voting stake in Scale AI, aiming to bolster Meta's AI capabilities [2] - The recruitment strategy included offering signing bonuses in the nine-figure range, with total compensation packages for key talents reaching up to $100 million [2][10] - By July 19, 2025, 40% of the 44 employees in the Meta Superintelligence Labs (MSL) came from OpenAI, 20% from DeepMind, and 15% from Scale AI, with salaries ranging from $1 million to $100 million [10] Employee Turnover - Following the recruitment freeze, at least nine employees from MSL have left within two months, including key figures who returned to previous employers [3][34] - The high compensation for new hires has led to feelings of marginalization among long-term employees, prompting several to resign [3][35] - Internal dissatisfaction has been reported regarding the management style and resource allocation, contributing to the turnover [35][36] Leadership and Team Structure - Alexandr Wang, the CEO of Scale AI, leads MSL, which has faced skepticism regarding his leadership due to his lack of deep AI background [4][9] - The MSL is organized into four teams: TBD Lab (focused on superintelligence), AI Products & Applied Research, MSL Infra (infrastructure), and FAIR (long-term exploratory projects) [32][33] - Key hires include Nat Friedman, Daniel Gross, and Yann LeCun, who bring extensive experience and strategic vision to the team [12][13][14][18] Challenges and Market Context - The recruitment freeze coincides with rising concerns about the sustainability of AI investments, as many projects fail to deliver quick returns [31][38] - Meta's AI assistant has seen low user engagement, with only 10% of monthly active users interacting with it, raising questions about the effectiveness of the current strategy [38] - The company is reportedly considering using models from Google Gemini or OpenAI to enhance its AI product capabilities amid internal challenges [38]