来源 | 硬AI 以下文章来源于硬AI ，作者硬ai 硬AI . AI时代，快人一步~ A I 竞 赛 升 级 M e t a 加 速 布 局 Meta首席执行官马克·扎克伯格正大幅提升其在人工智能领域的投入，此前Meta在AI发展方面已 落后于竞争对手。其大型语言模型Llama 4表现未如预期，旗舰模型"Behemoth"的发布也被推 迟。 蓝筹公司越来越多地依赖私人投资公司获取资金，以避免大型资本项目给自身资产负债表带来过 大压力。这些交易通常通过特殊目的载体或合资企业进行，资产管理公司获得大额少数股权，公 司则贡献资产以换取私人投资公司提供的资本。这种结构能帮助公司将类似债务的融资保留在资 产负债表之外，从而避免影响其杠杆率和信用评级。 社交媒体巨头Meta正寻求从PE巨头筹集高达290亿美元资金，以大规模建设其在美国的AI数据中 心，此举标志着该公司全面发力人工智能领域。 私人投资机构如阿波罗去年就与英特尔达成了一项110亿美元的协议，为英特尔在爱尔兰的半导 体制造工厂提供资金，以换取其业务部门的股权和有保障的现金流。阿波罗和黑石等资产管理公 司现已拥有或与主要保险公司和年金提供商建立了合作关系，这些 ...

Core Viewpoint - Meta is seeking to raise up to $29 billion from private equity firms to significantly expand its AI data center capabilities in the U.S. This move indicates a strong commitment to advancing in the artificial intelligence sector [2][4]. Group 1: Funding and Investment Plans - Meta is in advanced negotiations with several private equity giants, including Apollo Global Management, KKR, Brookfield, Carlyle, and Pimco, to raise $3 billion in equity and $26 billion in debt [2][4]. - This funding effort is expected to be one of the largest private financing deals in the sector [2]. Group 2: Competitive Landscape and Strategic Moves - CEO Mark Zuckerberg is increasing investments in AI, as Meta has previously lagged behind competitors. The company has faced challenges with its large language model Llama 4 and delayed the release of its flagship model "Behemoth" [4]. - To catch up, Meta announced a $15 billion investment in data labeling startup ScaleAI and recruited Scale's CEO to lead a new "superintelligence" team focused on general AI development [4]. - Meta's capital expenditure forecast for the year has been raised by up to 10%, now projected between $64 billion and $72 billion, due to additional investments in AI data centers and rising infrastructure hardware costs [4]. Group 3: Industry Trends - The trend of tech giants partnering with private equity firms to share investment risks and costs is becoming more common. For instance, OpenAI is collaborating with Blue Owl to invest $15 billion in a Texas data center [6]. - This approach allows companies to keep large capital projects off their balance sheets, thereby avoiding negative impacts on leverage ratios and credit ratings [6].

整理 | 华卫 近日，据外媒报道，Meta 平台公司已招募四名前 OpenAI 研究人员加入其新成立的超级智能实验 室。 消息称，此次招聘对象包括 2022 年加入 ChatGPT 开发团队的特拉皮特·班萨尔（Trapit Bansal）。 据悉，他在启动 OpenAI 强化学习项目中发挥了关键作用。强化学习作为一种 AI 训练方法，适用于 构建推理模型。 另外三名已加入 Meta 的 OpenAI 研究人员分别是卢卡斯·拜尔（Lucas Beyer）、亚历山大·科列斯尼 科夫（Alexander Kolesnikov）和翟晓华（Xiaohua Zhai）。据了解，这三人曾于去年底协助建立 OpenAI 苏黎世办公室，此前他们在谷歌母公司 Alphabet 旗下的 DeepMind 机器学习实验室工作。 此次招聘发生在 Meta 首次披露组建超级智能研究团队的数周后。该实验室将负责开发能在广泛任务 中超越人类表现的 AI 模型。据悉，Meta 成立该部门的背景是其内部开发的大型语言模型 Llama 4 Behemoth 面临性能问题——该模型于今年早些时候预览，但因性能担忧已推迟发布。 上周，OpenAI 透 ...

社交媒体巨头Meta正寻求从PE巨头筹集高达290亿美元资金，以大规模建设其在美国的AI数据中心，此 举标志着该公司全面发力人工智能领域。 据媒体28日援引知情人士透露，Meta与包括阿波罗全球管理公司、KKR、 Brookfield、Carlyle和Pimco 在内的多家PE巨头之间的谈判已进入深入阶段。 据了解，Meta计划从中募集30亿美元股权和260亿美元债务。公司目前正在讨论如何构建这笔巨额债务 融资，这有望成为同类私人融资中规模最大的交易之一。 AI竞赛升级，Meta加速布局 Meta首席执行官马克·扎克伯格正大幅提升其在人工智能领域的投入，此前Meta在AI发展方面已落后于 竞争对手。其大型语言模型Llama 4表现未如预期，旗舰模型"Behemoth"的发布也被推迟。 华尔街见闻此前提及，为追赶进度，Meta本月宣布向数据标注初创公司ScaleAI投资150亿美元，并聘 请Scale首席执行官Alexandr Wang加入新的"超级智能"团队，专注于通用人工智能的开发。扎克伯格本 人也在积极从竞争对手处招募AI人才，本周有三位顶尖OpenAI研究员宣布加盟Meta，此前有报道称扎 克伯格为工程 ...

Meta成功挖走OpenAI苏黎世办公室的三名核心创始研究员，此次挖角是扎克伯格"超级智能团队"组建计划的关键一步： 他不仅为顶尖人才开出超1亿美元薪酬包，还亲自通过WhatsApp联系目标研究员。有趣的是，OpenAI CEO奥特曼近日 刚刚公开嘲讽Meta的"撒钱"策略，宣称"我们最优秀的人才一个都没走"，并质疑Meta依赖金钱而非使命驱动的文化难以 吸引真正顶尖人才。 硬·AI 对于扎克伯格的魅力攻势，奥特曼在播客中表示："我真的很高兴，至少到目前为止，我们最优秀的人才都 没有决定接受他的那些提议。"他补充道，"我认为大量前期保证薪酬的策略，以及以此作为告诉某人加入的 理由，真正专注于此而不是工作和使命的程度，我不认为这会建立一个伟大的文化。" 不过，这一招聘策略产生了喜忧参半的结果。扎克伯格最近斥资140亿美元成功挖走了Scale AI的CEO Alexandr Wang，使这位28岁的年轻人成为科技行业历史上最昂贵的聘用之一。 然而，更大的目标仍然未能得手。包括OpenAI联合创始人Ilya Sutskever和John Schulman在内的重量级人 物都未被成功挖走，两人都继续联合创立了新的初创 ...

Core Insights - Meta is investing $10 billion to build the largest data center in the Western Hemisphere in Louisiana, covering an area equivalent to 1,700 football fields, to enhance its AI infrastructure and talent acquisition [1][3] - The company recently won a significant copyright case regarding its open-source AI model Llama, which is seen as a crucial factor in its competitive position in the AI race [3] - Meta's CEO Mark Zuckerberg has increased the annual capital expenditure range from $60 billion to $65 billion to $64 billion to $72 billion, emphasizing the transformative impact of AI on the company's operations [3][4] Investment and Talent Acquisition - Meta has invested $14.3 billion in AI startup Scale AI and plans to hire notable entrepreneurs, including former GitHub CEO Nat Friedman and Daniel Gross, to strengthen its AI capabilities [4] - The company's technology chief Andrew Bosworth highlighted the unprecedented valuation of tech talent in the current market, indicating a competitive landscape for acquiring skilled professionals [4] Competitive Landscape - Despite the significant investments, Meta's Llama 4 AI model received poor reviews from developers, causing the company to lag behind competitors like Google, OpenAI, and Anthropic in the AI foundational model race [5] - The release of Meta's flagship model Behemoth has been delayed, which is intended to serve as a "teacher" for new models within the mixed expert model architecture [5] Stock Performance and Market Sentiment - Meta's stock price has increased by over 40% in the past year, outperforming most tech companies, driven by investor optimism regarding the company's ongoing investments in AI [6] - A former Meta executive noted that while the stock performance appears strong, the company's AI efforts over the past year have been mediocre, suggesting a need for significant reform to maintain momentum [5]

Group 1 - Meta successfully recruited three core researchers from OpenAI, indicating the effectiveness of its aggressive hiring strategy led by CEO Mark Zuckerberg [1] - The recruited researchers, Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai, were previously responsible for establishing OpenAI's Zurich office and joined Meta's Superintelligence team [1] - Zuckerberg's recruitment strategy includes offering over $100 million compensation packages to attract top talent from competitors like OpenAI [2] Group 2 - OpenAI CEO Sam Altman acknowledged the high offers from Meta but expressed confidence that their best talent has not accepted these proposals [2] - Meta's recent hiring of Alexandr Wang, CEO of Scale AI, for $14 billion marks one of the most expensive hires in tech history, although it has not successfully recruited other key figures from OpenAI [2] - Meta faced setbacks in the AI field, particularly with the disappointing performance of the Llama 4 model, which led to internal and external criticism regarding its capabilities [3] Group 3 - The launch of Meta's large model "Behemoth" has been delayed, raising concerns within the leadership about its competitive edge compared to products from OpenAI, Anthropic, and Google [3] - Zuckerberg's ambition for Meta to have the best AI product by year-end has resulted in increased pressure on the AI team, leading to long hours and unmet expectations [3]

AI领域又有一笔百亿级的收购了。 6月10日，媒体曝光Meta将以149亿美元（折合人民币约1066亿元）的价格收购Scale AI 49%的股权。 这笔交易极为特殊——Meta不仅用现金回购老股，还将Scale CEO Alexandr Wang纳入麾下，出任新成立 的「超级智能小组」负责人。 按照股权比推算，此次交易Wang和团队有可能获得74亿美元。 这不仅是Meta为逆转AI困局押下的最大筹码，也标志着这位28岁、美国最年轻的白手起家亿万富翁， 正式站上AI权力的中心舞台。 01 一场150亿美金的"豪赌" 这次收购背后，透露着扎克伯格的"无奈"。 虽然Meta在AI领域的投入越来越多，但收效甚微。 先是核心旗舰大模型Behemoth"难产"。 今年4月，Meta在首届AI开发者大会上高调发布了其旗舰大模型Behemoth，号称参数规模高达2万亿， 是OpenAI与Anthropic的直接对手。 然而，模型性能未达预期，发布时间一再跳票。直到现在，这款"巨兽"依然没有对外上线，甚至可能要 等到今年秋季，甚至更晚。 大模型的落后，是扎克伯格所绝对不容许的。此前，扎克伯格在复盘移动互联网的时候，认为自己受 ...

Core Insights - Meta's CEO Mark Zuckerberg is personally assembling an elite team to accelerate the development of Artificial General Intelligence (AGI) due to disappointment with the company's performance in the AI sector [1][2][3] - The new team, referred to internally as the "superintelligence group," aims to position Meta as a leader in AGI, integrating its capabilities across all Meta products [1][2] - Meta plans to make a multi-billion dollar investment in Scale AI, which will be the largest external investment in the company's history, potentially raising Scale AI's valuation to $28 billion [2][3] Group 1: Team Formation and Leadership - Zuckerberg is directly involved in recruiting approximately 50 experts for the new team, including a new AI research head, and has adjusted the office layout for closer collaboration [1][2] - He has adopted a "founder mode" management style, reflecting a hands-on approach to address dissatisfaction with Meta's latest AI models, particularly Llama 4 [2][3] - The establishment of a WhatsApp group named "Recruiting Party" among executives indicates a continuous discussion on talent acquisition [3] Group 2: Investment and Market Position - The investment in Scale AI is expected to enhance Meta's AI capabilities, with the founder of Scale AI likely joining the superintelligence group post-transaction [2][3] - Zuckerberg believes that Meta's strong cash flow from its advertising business allows for significant investments in AI, with planned capital expenditures reaching hundreds of billions [3] Group 3: Challenges and Internal Dynamics - Meta has delayed the release of its "Behemoth" model due to concerns about its limited improvements, reflecting internal pressures and performance expectations [3][4] - The company has faced criticism regarding the performance of its Llama models, with internal and external stakeholders questioning their effectiveness [4][5] - There is ongoing internal turmoil, with high-level departures attributed to a lack of focus on AGI's disruptive potential, as well as challenges from competitors in the AI space [5][6]

Meta的交易将Scale AI的估值从去年的138亿美元推高至280亿美元，凸显了扎克伯格在AI军备竞赛中的焦虑与野心。对 于Meta而言，收购Scale AI不仅能获得关键的数据标注能力，更重要的是获得Wang及其团队对竞争对手AI模型构建方式 的深度了解。 硬·AI 作者 | 赵雨荷 值得注意的是，Meta选择收购49%股份而非全资收购，这一结构设计显然旨在规避更严格的监管审查。 目前Meta正因收购Instagram和WhatsApp面临反垄断调查，全资收购可能引发监管部门审查。 这笔交易凸显了扎克伯格在AI军备竞赛中的焦虑与野心，也反映了当前AI领域的激烈竞争态势。谷歌此前 投资Character.AI，微软重金押注Inflection AI，科技巨头们都在通过大额投资而非直接收购的方式，获得 顶尖AI人才和技术。 01 Scale AI： AI训练数据"独角兽" Scale AI成立于2016年，专注于提供大规模数据标注服务——也就是为ChatGPT等高级AI工具提供经过精 心整理的训练数据。 该公司去年估值为138亿美元， 2024年营收约8.7亿美元，预计今年将突破20亿美元大关。但公司仍处于 ...