GOOGL：美银看好 AI 货币化上行空间，上调 TPU 出货预期 美银重申对谷歌（GOOGL）的 "买入"评级，强调 Alphabet 推出原生 AI 模式广告格式（包括 "Direct Offers"）标志 着向以代理型、交易驱动的货币化转变，转化率、意向捕捉和衡量的改善支持搜索业务在 2026 年的上行。 GOOGL：美银看好 AI 货币化上行空间，上调 TPU 出货预期 美银重申对谷歌（GOOGL）的 "买入"评级，强调 Alphabet 推出原生 AI 模式广告格式（包括 "Direct Offers"）标志 着向以代理型、交易驱动的货币化转变，转化率、意向捕捉和衡量的改善支持搜索业务在 2026 年的上行。 另据供应链反馈，美银将 2026 年 TPU（张量处理单元）出货量预估上调至 46 另据供应链反馈，美银将 2026 年 TPU（张量处理单元）出货量预估上调至 460 万个单位（此前为 400 万个；2025 年为 230 万个），称内部工作负载强劲，且在 Meta 的 MTIAASIC 项目因管理层变动而推迟的情况下，外部需求 有所增加；分析师还认为谷歌正在加速部署基于 ARM 的服务器 ...

此外，衡量需求强劲程度的另一项指标——新债溢价(即投资者购买新债相比该公司现有证券所获得的 额外收益率，也称新发行折让)今年平均仅为0.02个百分点，即2个基点。去年这一数字为3.3个基点。 截至本周三，美国高评级证券的风险溢价(即利差)今年已收窄0.03个百分点至0.75个百分点，接近数十 年来的最低水平。Wealth Enhancement Group投资组合咨询总监Ayako Yoshioka表示，如此高的估值几乎 没有进一步改善的空间，她对信贷市场持中性展望。 她表示："我们没有看到任何看跌的催化剂，但考虑到利差目前的水平，看涨的空间也非常有限。" 智通财经APP注意到，据LSEG Lipper数据显示，在截至周三的一周内，美国债券投资者再次向高评级 债券基金注资43亿美元，这已是连续第11周录得资金流入。随着投资者争相买入仍能提供可观收益率的 债券，此类资金流入趋势持续强劲。 在 1月份投资级债券基金创下433亿美元(五年来最大月度流入量)的纪录后，近期资金继续涌入短期和中 期投资级债券基金。持续的资金流入为今年企业债发行市场的需求提供了有力支撑。 2026年至今，高评级公司已发行约3090亿美元的 ...

Core Insights - The recent departures of xAI co-founders Yuhuai Tony Wu and Jimmy Ba have sparked significant industry discussion, signaling a potential shift towards smaller, AI-driven teams redefining innovation in the sector [1][2] - The trend of key personnel leaving established AI companies like OpenAI to pursue entrepreneurial ventures is becoming a notable pattern in the industry, indicating a movement from large organizations to startups [3][4] Group 1: xAI Developments - xAI's founding team has halved since its inception in 2023, with several core technical figures departing, which may impact the company's future capabilities and direction [3] - Wu's and Ba's statements reflect a broader trend in the AI industry, emphasizing the potential of small teams leveraging AI technology to create impactful solutions [2][3] Group 2: OpenAI Talent Exodus - A significant number of key personnel from OpenAI have left to establish their own startups, focusing on various aspects of AI, including safety, general intelligence systems, and AI search [4][5] - Notable startups emerging from this talent exodus include Safe Superintelligence, Thinking Machines Lab, and Perplexity AI, each targeting different niches within the AI landscape [7][8][10] Group 3: Investment and Valuation Trends - Safe Superintelligence has raised approximately $10 billion in funding, achieving a valuation of around $50 billion, with further funding rounds increasing its valuation to about $320 billion [7] - Thinking Machines Lab has also attracted significant investment, securing $20 billion in seed funding and reaching a valuation of approximately $120 billion [9] - Perplexity AI has gained traction as an early AI search tool, supported by investments from notable figures and firms, including Jeff Bezos and Nvidia [11] Group 4: Competitive Landscape - Anthropic, founded by former OpenAI employees, is focusing on large model development and has achieved a valuation of $615 billion following its E-round funding [14] - Character.AI, co-founded by former Google Brain researchers, has become a leader in AI virtual character interactions, boasting over 20 million monthly active users and a valuation of around $10 billion [26][27] Group 5: Future Outlook - The AI industry is evolving from a focus on foundational model breakthroughs to practical applications and long-term strategic planning, with a clear trend towards safety and system architecture [28] - The emergence of open-source ecosystems is enabling smaller teams and individual developers to redefine the execution capabilities of AI, suggesting a dynamic future for the industry [29]

Group 1 - The State Administration for Market Regulation opposes manufacturers engaging in "involution" and selling cars at a loss, aiming to promote healthy competition in the automotive market [3] - The newly released guidelines clarify the boundaries of pricing behavior in the automotive industry, encouraging compliance and fair competition among manufacturers and sellers [3] - The guidelines specify that automotive manufacturers must not engage in pricing behaviors that aim to eliminate competitors or monopolize the market, which poses significant legal risks [3] Group 2 - OpenAI announced the release of a new real-time programming model, GPT-5.3-Codex-Spark, which is a more compact version of GPT-5.3-Codex, optimized for low-latency hardware [5][7] - The Codex-Spark model can generate over 1000 tokens per second and is currently available as a research preview for ChatGPT Pro users [5][7] - The model operates on Cerebras' Wafer Scale Engine 3, which features 40 trillion transistors and 900,000 AI cores, providing extremely low latency [7] Group 3 - Multiple automotive companies, including Geely, Chery, and BYD, have disclosed their plans for solid-state battery technology, with Geely aiming for small-scale production by 2027 and a long-term goal of mass production by 2030 [18][19][20] - Chery plans to start pilot production of solid-state batteries by 2026 and aims to validate the technology in real vehicles by 2027 [20] - BYD is focusing on sulfide solid-state batteries, expecting to achieve small-scale production by 2027 [23] Group 4 - Meizu's smartphone business is reportedly facing dissolution, with the Meizu 23 project allegedly halted [25][26] - The company has not officially responded to the rumors, and customer service has stated that no notifications regarding the business's status have been received [26] Group 5 - ByteDance is set to release the upgraded Doubao model on February 14, 2026, which will include significant enhancements to its foundational model capabilities [34][35] - The Doubao 2.0 model is expected to be a multimodal model with 1 trillion parameters, marking a substantial advancement in the company's AI capabilities [35] Group 6 - Xiaomi has launched its first-generation robot VLA model, which features 4.7 billion parameters and excels in visual language understanding and real-time execution capabilities [39][42] - The model has achieved state-of-the-art results in various benchmarks and is designed for high efficiency in real-world tasks [42] Group 7 - Lenovo reported a 72% year-on-year increase in AI-related revenue, which now accounts for 32% of the company's total revenue, driven by growth in AI PCs, smartphones, and servers [32][36] - The company has successfully navigated challenges in the global AI supply chain, maintaining its commitment to double-digit revenue growth and profitability [32]

（原标题：?"Gemini热潮"再度席卷全球! 谷歌(GOOGL.US)Deep Think"硬核升级"直指大型科研工程） 智通财经APP获悉，美国科技巨头谷歌(GOOGL.US)对其风靡全球的Gemini 3 AI大模型的Deep Think(深 度思考)模式在科学、编程、研究与工程学等方面进行了重大升级，这一最新升级聚焦于解决现代科学 研究与工程领域的诸多复杂挑战，可谓引爆新一轮席卷全球的"Gemini AI狂热浪潮"。谷歌官方在一份 声明中表示，新的Deep Think模式现已在Gemini 3系列AI应用产品中面向 Google AI Ultra 订阅用户开 放。该公司补充称，这是其首次通过Gemini API向部分研究人员、工程师及大型企业提供Deep Think功 能。 据谷歌介绍，通过更新后的Gemini 3 AI大模型Deep Think(深度思考)模式，订阅用户们可以将草图变为 可3D打印的现实物体。Deep Think拥有强大的能力分析图纸、对极度复杂形状建模，并生成可用于3D 打印制作实体物品的文件。 该公司在周四的一篇博客文章中表示："除了其最先进的性能之外，Deep Think还旨在 ...

办公室的灯刚熄，家里的灯又亮起。 一天结束，也是他真正开始的一刻。 Demis Hassabis 在《财富》最新视频采访里透露了他的作息： "大约晚上 10 点，我会开始第二轮工作，一直做到凌晨 4 点。" 白天，他的会议一个接一个，几乎没有空隙；夜里，他留出六小时，只做一件事：思考。 这种作息，他已经坚持了十年。而现在的 AI 行业，让这种节奏显得愈发必要。 他领导的 Google DeepMind，正处于一个关键时刻： "我们进展飞快。" 他的语气很平静，但话里藏着整个行业的焦虑：技术越接近临界点，洗牌的速度就越快。 采访里有一句话非常关键： "领先可能只保持几个月。" 这就是当下的 AI 行业。顶尖实验室之间的差距越来越小，领先优势随时会被打破。 在这场采访里，Hassabis 谈了竞争、泡沫、算力与人才，也推演了 AI 将如何重塑科学、医疗与未来的 设备形态。 第一节｜洗牌已开始：竞争在加速 从竞争开始。 模型竞赛：更新速度决定位置 Hassabis 对 Gemini 3 的表现很满意，但他也坦白，竞争从未像现在这样激烈。 因为所有人都在冲刺：模型更新从"一年一版"变成"几个月一版"，新能力从单点突破 ...

来源：汲古科技 Google DeepMind 团队表示，该模型旨在弥合深奥科学理论与日常工程应用之间的鸿沟。 在具体的性能评估中，Gemini 3 Deep Think 展现出了亮眼的数据。 Gemini 3 Deep Think 发布：一张草图直接获得3D模型 Google 今日宣布对 Gemini 3 Deep Think 进行重大升级。这是一款专为解决科学、研究和工程挑战而设 计的推理模型。 目前，Google AI Ultra 订阅用户已可以在 Gemini App 中使用该更新版本，同时，Google 首次向选定的 研究人员、工程师和企业开放了 Gemini API 的 Deep Think 早期访问权限。 从抽象理论到硬核工程 此次更新的核心在于"深度思考"能力的进化。Gemini 3 Deep Think 不仅强化了科学知识库，更着重于解 决那些缺乏明确边界、数据混乱或不完整的复杂问题。 在测试现代前沿模型极限的"Humanity's Last Exam"基准测试中，该模型在不使用额外工具的情况下达 到了 48.4% 的准确率，树立了新的行业标准。 更为引人关注的是其在 ARC-AGI-2 测 ...

炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ 来源：汲古科技 Gemini 3 Deep Think 发布：一张草图直接获得3D模型 目前，Google AI Ultra 订阅用户已可以在 Gemini App 中使用该更新版本，同时，Google 首次向选定的 研究人员、工程师和企业开放了 Gemini API 的 Deep Think 早期访问权限。 从抽象理论到硬核工程 此次更新的核心在于"深度思考"能力的进化。Gemini 3 Deep Think 不仅强化了科学知识库，更着重于解 决那些缺乏明确边界、数据混乱或不完整的复杂问题。 Google DeepMind 团队表示，该模型旨在弥合深奥科学理论与日常工程应用之间的鸿沟。 在具体的性能评估中，Gemini 3 Deep Think 展现出了亮眼的数据。 Google 今日宣布对 Gemini 3 Deep Think 进行重大升级。这是一款专为解决科学、研究和工程挑战而设 计的推理模型。 在测试现代前沿模型极限的"Humanity's Last Exam"基准测试中，该模型在不使用额外工具的情况下达 到了 48.4% 的 ...

美国科技巨头谷歌(GOOGL.US)对其风靡全球的Gemini 3 AI大模型的Deep Think(深度思考)模式在科学、编程、研究与工程学等方面 进行了重大升级，这一最新升级聚焦于解决现代科学研究与工程领域的诸多复杂挑战，可谓引爆新一轮席卷全球的"Gemini AI狂热 浪潮"。谷歌官方在一份声明中表示，新的Deep Think模式现已在Gemini 3系列AI应用产品中面向 Google AI Ultra 订阅用户开放。该 公司补充称，这是其首次通过Gemini API向部分研究人员、工程师及大型企业提供Deep Think功能。 据谷歌介绍，通过更新后的Gemini 3 AI大模型Deep Think(深度思考)模式，订阅用户们可以将草图变为可3D打印的现实物体。Deep Think拥有强大的能力分析图纸、对极度复杂形状建模，并生成可用于3D打印制作实体物品的文件。 该公司在周四的一篇博客文章中表示："除了其最先进的性能之外，Deep Think还旨在推动实际应用，使研究人员们能够深度解读复 杂数据，并使工程师们能够通过代码与编程对复杂物理系统进行建模。"最重要的是，我们正在努力将Deep Think ...

Investment Rating - The report assigns an "Outperform" rating for the industry, indicating expected performance above the market benchmark by more than 10% over the next 6 to 12 months [33]. Core Insights - Major tech companies are significantly increasing their investments in AI for 2026, focusing on talent acquisition, computing power infrastructure, and marketing expenditures. The competition for consumer-facing AI Agent products is expected to intensify during the Spring Festival period in China [2]. - The report emphasizes the transformative potential of AI in changing user interactions with software and the internet, highlighting the value creation from large models. It suggests focusing on companies with the most certainty in computing power and large models, including Alibaba, Baidu, and Tencent [2]. Summary by Sections Company Developments - SpaceX has fully acquired xAI, with a post-merger valuation of $1.25 trillion. The merger will allow xAI to operate as a subsidiary of SpaceX, integrating its Grok model with SpaceX's satellite data [17]. - Meta has launched a series of AI advertising tools, including AI Video Generation 2.0, which simplifies the creative production process for advertisers [20]. - Kunlun Technology has released the Skywork desktop version, a local multi-model AI office agent that prioritizes data security and ease of use [21]. - OpenAI has launched two core products, including an upgraded programming model and an enterprise-level AI platform, enhancing its commercial capabilities [22]. - Meta is testing a standalone AI video application called Vibes, focusing on AI-generated content [23]. - Four major tech giants (Google, Amazon, Meta, Microsoft) plan to invest a total of $610 billion in AI infrastructure, representing over 75% of global investment in this area [24]. Underlying Technologies - Step 3.5 Flash model by Jumpshare has been released, utilizing a sparse MoE architecture to address computing power challenges [25]. - Alibaba's Tongyi Qianwen team has open-sourced the Qwen3-Coder-Next programming model, achieving high performance with low computational costs [26]. - Anthropic has released Claude Opus 4.6, significantly expanding its context window to 1 million tokens, enhancing its capabilities for complex document analysis [26]. - The Chinese Academy of Sciences has introduced the "Feiyu-1.0" model, focusing on dual-direction coupling technology for environmental data analysis [27]. Industry Policies - The Ministry of Industry and Information Technology has issued a notice to enhance AI computing power infrastructure, aiming to establish a national interconnected computing power node system [28]. Investment Recommendations - The report suggests monitoring companies with strong positions in computing power and large models, such as Alibaba, Baidu, and Tencent, as they are expected to lead in the rapidly evolving AI landscape [29].