记者丨彭新 编辑丨包芳鸣 美国时间12月8日，谷歌举办直播活动，展示其AI眼镜生态进展，并宣布首款AI眼镜产品将于 2026年上市。 抢占A I新入口 目前来看，谷歌此次深度布局AI眼镜，与其说是进军，不如说是"重返"。 2012年，谷歌推出AR（增强现实）眼镜Google Glass（谷歌眼镜），向全球普及AR概念，然 而这一产品并不成功。第一代谷歌眼镜1500美元的高昂售价、极为有限的功能和过短的续航 时间等，使其在2014年量产进军消费市场时出师不利，迅速成为行业过客。 经历了十年前在C端市场的折戟沉沙后，谷歌此番重返头戴式智能设备赛道，战略已发生根本 变化，即从单纯的头戴式摄像机转向AI原生交互入口，并在生态系统层面做出诸多努力，包 括打造Android XR平台、大力投资发展AI大模型Gemini等。 谷歌介绍，在Gemini加持下，Android XR设备具备了强大的视觉理解能力，可以理解用户看 到的环境，并可以智能执行任务，带来更高效、便捷的使用体验。作为专门为智能时代打造 的操作系统平台，Android XR将成为驱动XR生态的核心引擎，支持头显、智能眼镜等多样化 形态的硬件。 "为了让AI与 ...

12月9日凌晨，最近春风得意的 Google 举办了一场 Android Show: XR，第一次明确了 Android XR 的设备路线，第一次正式展示与三星合作开发的 AI 眼镜原 型机。 但如果只看到这些，这场 Android Show: XR 很容易被低估。 没有价格、没有参数堆叠，也没有试图用一款「颠覆性设备」占据活动核心。整场活动更像是一份克制到近乎冷静的路线说明，Google 花了大部分时间强 调 Gemini 与 Android XR 的系统整合，讲开发框架、讲 API，讲「Android XR 将覆盖哪些设备形态」，避免把焦点压在某一款终端硬件上。 在官方叙事中，Android XR 已经不再只是「支持头显的 Android 版本」，更是一个以 Gemini 为核心的全新计算平台。 低调的发布会，高调的XR新蓝图 Google 强调，Gemini 将作为默认的智能层存在于 Android XR 中，贯穿视觉、语音、环境感知与交互理解；而 Android XR 的角色，则是为这种多模态 AI 提 供稳定、可扩展、跨设备体验的系统载体。 以此为核心，Google 也一次性铺开了四条产品路线： X ...

Core Viewpoint - Google currently holds a cost advantage in AI training with its TPU chips, operating at a negative 30% profit margin, which allows it to suppress competitors. However, this advantage is expected to reverse with the introduction of NVIDIA's Blackwell chip cluster in early 2026, potentially reshaping the competitive landscape of the AI industry [1][4][11]. Group 1: Cost Structure and Competitive Dynamics - Gavin Baker highlights that Google's TPU chips are akin to "fourth-generation jet fighters," while NVIDIA's Hopper chips are compared to "World War II P-51 Mustangs," indicating a significant cost advantage for Google [4]. - The transition from NVIDIA's Hopper to Blackwell is described as one of the most complex product transformations in tech history, with substantial increases in data center rack weight and power consumption [5]. - Baker anticipates that the first models trained on Blackwell will debut in early 2026, with xAI playing a crucial role in NVIDIA's deployment strategy [6]. Group 2: Supply Chain and Design Strategy - Google's conservative design choices and supply chain strategy may limit its long-term competitiveness, as it outsources backend design to Broadcom, incurring significant costs [7]. - The estimated annual payment to Broadcom could reach approximately $15 billion by 2027, raising questions about the economic rationale behind this outsourcing [7]. - The introduction of MediaTek as a second supplier is seen as a warning to Broadcom, but this diversification may slow down TPU's development pace compared to NVIDIA's rapid GPU iterations [9][10]. Group 3: Strategic Implications - Once Google loses its status as the lowest-cost producer, its strategic computing approach will fundamentally change, making it challenging to maintain a negative profit margin [11]. - The shift in cost dynamics with the Blackwell cluster moving towards inference applications could lead to significant financial strain for Google, potentially impacting its stock performance [11]. - Baker emphasizes that the gap between NVIDIA's GPUs and Google's TPUs will widen further with the release of the next-generation Ruben chip [12].

Core Viewpoint - OpenAI has initiated a "red code" alert in response to increasing competition from Google, leading to a strategic shift that prioritizes short-term commercial goals over the long-term vision of achieving Artificial General Intelligence (AGI) [1][3][4]. Group 1: Strategic Shift - OpenAI has decided to pause long-term projects, including the Sora video generator, for eight weeks to focus on improving user engagement with ChatGPT [5][6]. - The company aims to leverage user signals to enhance ChatGPT's performance on model rankings and increase user retention [3][5]. - This decision reflects an internal struggle between the commercialization team, advocating for immediate product success, and the research team, which prefers pursuing cutting-edge technological breakthroughs [5][6]. Group 2: Competitive Landscape - OpenAI faces significant challenges as Google's recent launches, such as the Nano Banana image generator and Gemini 3 model, have quickly gained market traction and outperformed OpenAI's offerings [4][8]. - The financial sustainability of OpenAI is under pressure, especially with a recent $1.4 trillion infrastructure contract that may be difficult to fulfill if user growth slows [4][8]. - OpenAI's valuation reached $500 billion, with weekly active users exceeding 800 million, necessitating a robust user growth strategy to support its operational costs [8]. Group 3: User Engagement Strategy - The "user signal" strategy, which relies on user feedback for model training, has led to high engagement but raised concerns about the potential negative impact on user mental health [9][10]. - OpenAI previously faced backlash for prioritizing user engagement over safety, leading to a temporary shift in training methods to address these issues [9][10]. - The company plans to release a new model in January that will improve image quality, speed, and personalization, marking the end of the "red code" state [10][11]. Group 4: Future Outlook - OpenAI is navigating the delicate balance between immediate commercial success and the long-term goal of AGI, similar to challenges faced by other tech giants [10][11]. - The company must find a way to manage high operational costs while addressing ethical concerns related to user safety and mental health [11].

Core Viewpoint - Google holds a cost advantage in AI training due to its TPU chips, compared to Nvidia's Hopper chips, which are considered outdated [1] Group 1: Cost Advantage - Gavin Baker highlights that Google's TPU chips provide a low-cost advantage in the AI training sector, likening them to "fourth-generation jet fighters" [1] - In contrast, Nvidia's Hopper chips are compared to "World War II-era P-51 Mustangs," indicating a significant technological gap [1] - This cost advantage allows Google to operate its AI business at a negative profit margin of 30%, effectively "sucking the economic oxygen out of the AI ecosystem" [1] Group 2: Future Competition - Baker warns that this situation may change with the introduction of Nvidia's Blackwell chip cluster in early 2026, which will enhance training capabilities [1] - The subsequent release of the more easily deployable GB300 chip is expected to further shift the competitive landscape [1] - If Google loses its cost advantage, it could reshape the competitive dynamics and economic models within the AI industry [1]

Group 1 - The core viewpoint of the article highlights the tightening competitive landscape in the advertising platform industry, with several emerging platforms rapidly catching up to industry leaders [1][2] - In the gaming sector, Apple Ads continues to lead the global iOS gaming market, while AppLovin is gaining ground, particularly in North America and Western Europe, with Mintegral ranking third in terms of scale [1][2] - Google Ads maintains its dominance in Android gaming advertising, with strong growth from platforms like Mintegral and adjoe [1][2] Group 2 - In the non-gaming sector, Apple Ads remains the top player in iOS non-gaming advertising, followed by Meta Ads, TikTok for Business, Google Ads, and Snapchat [2] - Google Ads leads the Android non-gaming advertising strength rankings, with Meta Ads and TikTok for Business following closely [2] - The report indicates a shift in advertising spending, with a concentration of budgets towards leading platforms; 60% of the top five media channels experienced annual spending growth, while only 30% of channels ranked 11th to 20th saw growth [2] Group 3 - The introduction of a new creative index in the report allows for the measurement of advertising platform effectiveness from a creative perspective, marking a significant innovation [3] - The report suggests that the era of distributed competition has arrived, with a more balanced budget allocation among advertisers, leading to a decline in budget share for Google and Meta [3] - Mintegral has made significant strides, moving up three places in the iOS advertising strength rankings and entering the top five, while also debuting on e-commerce rankings [3] Group 4 - The acceleration of AI integration is transforming the advertising industry, with AI becoming a new productive force rather than just a tool, leading to potential automation of the entire marketing process [4] - The non-gaming user segment is identified as the main growth battlefield, with non-gaming growth outpacing that of gaming, particularly in areas like short videos, entertainment, fintech, and AI tools [4] - Future competitive advantages will depend on how well companies can integrate AI into their organizational structures and processes [4]

Core Insights - Nvidia's next-generation Blackwell chips and subsequent products are expected to reshape the cost structure of AI training, potentially ending Google's TPU cost advantage [1] - The transition from Nvidia's Hopper to Blackwell is one of the most complex product transformations in tech history, creating an unexpected advantage window for Google [2] - Google's conservative design choices and supply chain strategies in TPU development may limit its long-term competitiveness [4][5] Group 1: Nvidia's Blackwell Chips - The Blackwell chip cluster is set to begin training use in early 2026, with the GB300 chip following, which will be easier to deploy [1][2] - The first models trained on Blackwell are expected to be launched by xAI in early 2026 [2] - The GB300 chip will feature "plug-and-play" compatibility, allowing for direct replacement of existing GB200 infrastructure without additional modifications [3] Group 2: Google's TPU Challenges - Google's TPU architecture decisions, including outsourcing backend design to Broadcom, may result in significant annual payments, limiting profitability [4] - The introduction of MediaTek as a second supplier signals a warning to Broadcom, but this diversification may slow down TPU development [5] - If Google loses its status as the lowest-cost producer, its strategic computing approach will fundamentally change, making it difficult to maintain a negative profit margin [6]

Summary of Key Points from the Conference Call Industry Overview - The conference call focuses on the **Global Semiconductors, Hardware, Internet & Software** industry, particularly the **AI server market** and the **AI supply chain** dynamics [1][2]. Core Insights and Arguments 1. **Investment in Data Centers**: Total investment in upcoming and under-construction data centers is approximately **US$840 billion**. Major projects include OpenAI's agreements with semiconductor and cloud service providers (CSPs) [3][38]. 2. **AI Capital Expenditure (Capex)**: Following 3Q25 earnings, consensus estimates for 2026 capex for major CSPs have been raised by nearly **20%**, projecting total capex to grow at a **36% CAGR** to around **US$630 billion** from 2024 to 2027 [3][37]. 3. **Server Shipment Growth**: Global server and high-end GPU AI server shipments are expected to grow at **3%** and **31% CAGR**, respectively, from 2025 to 2027. High-end GPU server shipments are projected to increase by **46%** in 2026 [4][58]. 4. **ASIC and GPU Demand**: ASIC adoption is increasing, with projections indicating that ASICs will comprise **47%** of total CoWoS-based AI chip shipments in 2026. The demand for TPUs is also expected to grow by approximately **75% YoY** [4][5]. 5. **Financial Performance of Suppliers**: ASIC supply chain stocks have recently outperformed GPU supply chains, driven by significant orders from Broadcom and successful product launches from OpenAI and Google [6][34]. Additional Important Insights 1. **Circular Financing Concerns**: There are concerns regarding "circular financing" among AI giants, particularly with OpenAI's substantial commitments to various suppliers, which may lead to a "too big to fail" scenario [34][35]. 2. **OpenAI's Revenue Projections**: OpenAI's revenue for 1H25 was reported at **US$4.3 billion**, with a target of **US$13 billion** for the full year. However, the company has significant purchase commitments exceeding **US$1 trillion** over the next 5-7 years [35][36]. 3. **Major CSPs' Capex Guidance**: Companies like Microsoft, Amazon, and Google have increased their capex guidance significantly for 2025, indicating a strong focus on AI infrastructure [57]. 4. **Emerging Neoclouds**: Neoclouds are gaining traction with flexible "build-to-order" models, showcasing a growing revenue backlog and partnerships with major tech firms [43]. Investment Ratings and Price Targets - **Chroma**: Rated Outperform, Price Target (PT) = NT$830 [11] - **Delta**: Rated Outperform, PT = NT$1190 [12] - **Unimicron**: Rated Outperform, PT = NT$220 [13] - **Quanta**: Rated Underperform, PT = NT$250 [14] - **Google**: Rated Market-Perform, PT = $305 [15] - **Meta**: Rated Outperform, PT = $870 [16] - **Amazon**: Rated Outperform, PT = $300 [17] - **Microsoft**: Rated Outperform, PT = $645 [18] - **AMD**: Rated Market-Perform, PT = $200 [19] - **NVIDIA**: Rated Outperform, PT = $275 [21] - **TSMC**: Rated Outperform, PT = NT$1,444 [25] This summary encapsulates the key points discussed in the conference call, highlighting the significant trends, financial projections, and investment opportunities within the AI server and semiconductor industries.

Core Insights - Major tech giants are collaborating to establish a unified infrastructure for artificial intelligence, aiming to break down interoperability barriers between AI agents and enterprise applications [1][2] - The newly formed "Agentic Artificial Intelligence Foundation" will focus on developing open-source standards to ensure AI agents can work seamlessly across platforms [1][3] - This initiative is likened to the establishment of interbank electronic payment standards, facilitating smoother integration of AI automation into enterprise software [1][2] Industry Collaboration - The foundation will be organized by the Linux Foundation, indicating a rare collaboration among competitors in the tech industry at the foundational protocol level [1][3] - The collaboration is seen as a positive signal for industry maturity, as tech giants choose to work together to expand the market rather than compete [3] Standardization Efforts - The foundation will initially focus on standardizing three existing open-source AI tools, including connection protocols, instruction formats, and local running agents [4] - Key tools include: - Model Context Protocol (MCP) developed by Anthropic, which standardizes communication between AI models and APIs [4] - Agents.md by OpenAI, which standardizes the instruction format for coding agents [4] - Goose by Block, an open-source AI agent that can run locally without internet dependency [4] Security Challenges - Despite the acceleration of standardization, enterprise applications face significant security challenges, particularly concerning "prompt injection attacks" [6] - Companies are increasingly aware of the risks associated with connecting AI agents to operational tools, emphasizing the need for improved security measures [6] - The newly established foundation aims to address these security and technical collaboration issues to pave the way for large-scale commercialization of AI agents [6]

Market Overview - The Hang Seng Index has short-term support at the 25,000 point level, with recent hawkish signals from the Federal Reserve indicating limited rate cut space in 2026. Economic conditions in mainland China are cooling, and corporate earnings in Hong Kong are unlikely to improve significantly in the short term [2][3] - The Hang Seng Index closed at 25,434, down 1.29% year-to-date, with a cumulative increase of 26.79% [5] Sector Focus - The AI sector is gaining attention with the launch of AI glasses and smartphones, benefiting related stocks [7] - The biopharmaceutical sector is expected to thrive due to rising flu cases and favorable financing conditions [7] - The insurance sector is seeing improved investment returns driven by strong A-share performance [7] Corporate News - Vanke (2202) is reportedly willing to pay some interest to bondholders to facilitate debt extension negotiations [10] - Xiaomi (1810) is undergoing personnel adjustments in China and plans to close unprofitable stores [10] - Longi Green Energy (6869) is raising over HKD 2.2 billion through a nearly 15% discounted share placement [10] - SenseTime (0020) reports double-digit growth in domestic chip computing power and plans to launch a new generation of AI models in spring [10] - Novartis (NVS) has reached a USD 1.7 billion drug target collaboration agreement with UK biotech company Relation [10] - Midea (0300) has completed its share repurchase plan, buying back 135 million A-shares for approximately CNY 10 billion [10] Economic Indicators - The U.S. job openings reached 7.67 million in October, the highest in five months, alleviating concerns about labor market deterioration [8] - China's urban rail transit completed a passenger volume of 2.83 billion in November, showing a 4.4% year-on-year increase [8] - The Chinese express delivery development index for November was 478.1, up 3% year-on-year, indicating a positive trend in the industry [8] Investment Sentiment - Approximately half of the surveyed companies in Hong Kong are optimistic about the economic outlook for next year, a significant increase from 18% last year [9] - The Hong Kong government is planning to implement an automatic exchange of tax information related to cryptocurrency transactions starting in 2028 [9]