IT之家 1 月 21 日消息，据路透社报道，Facebook 母公司 Meta 首席技术官于当地时间周三表示，该公司新成立的人工智能实验室已于本月在内部交付了 首批人工智能模型。 不过他同时称，Meta 自 2025 年起在人工智能领域的一系列重磅投入，如今已开始显现可观的回报。 此前，Meta 首席执行官马克・扎克伯格对公司人工智能业务领导层进行了重大改组，组建了新实验室，并以高薪挖揽人才，希望助力公司在竞争白热 化的人工智能技术前沿领域占据优势。外界正密切关注 Meta 在该领域的发力动向。 此前，Meta 的 Llama 4 因性能问题饱受诟病，而与此同时，Alphabet 旗下的谷歌等竞争对手，在这场颠覆性、高回报的人工智能竞赛中已抢占先机。 博斯沃思表示，相关技术的研发目前尚未完成。 在谈及人工智能模型的研发周期时，博斯沃思指出："模型训练完成后，还有大量工作亟待推进，要把模型优化到能够真正在公司内部投入使用、并面 向消费者开放的程度。" 在达沃斯世界经济论坛年会间隙举行的新闻发布会上，Meta 首席技术官安德鲁・博斯沃思称，其于去年组建的"Meta 超级智能实验室"团队研发的这批 模型展现出巨大 ...

根据《金融时报》在去年12月的报道，Meta正在研发一款代号为"Avocado"的文本AI模型，计划在第一 季度发布。该公司同时还在开发一款聚焦图像和视频的模型，代号为"Mango"。博斯沃思并未具体说明 哪些模型已经在公司内部交付。 Meta的AI进展正受到业界密切关注。此前，Meta CEO马克.扎克伯格(Mark Zuckerberg)采取重大举措改 组AI领导层、组建新实验室并以天价薪酬争夺人才，希望能在竞争白热化的技术前沿领域取得胜利。 (作者/箫雨) 博斯沃思周三在达沃斯世界经济论坛年会间隙举行的一次新闻发布会上表示，去年成立的Meta超级智 能实验室团队打造的模型展现出了很大潜力。 "他们基本上才刚开展工作大约六个月，甚至还不到。"博斯沃思说道。他补充说，该团队的AI模型"非 常出色"。 Meta CTO博斯沃思 凤凰网科技讯北京时间1月21日，据路透社报道，Meta首席技术官安德鲁.博斯沃思(Andrew Bosworth)周 三表示，公司新成立的AI实验室本月已在内部交付了首批重磅AI模型。 ...

Group 1 - "The Magnificent Seven" refers to a group of tech stocks that have consistently impressed investors, particularly in the area of artificial intelligence (AI) [1][2] - These stocks have significantly contributed to the S&P 500's performance in recent years, indicating ongoing positive momentum [2] - Despite their gains, some of these companies, particularly Meta Platforms, are trading at reasonable valuations, with Meta being highlighted as the cheapest among them [3][5] Group 2 - Meta Platforms is currently trading at 20 times forward earnings estimates, which is lower than its peers in the Magnificent Seven [5] - The company has a strong presence in social media, with approximately 3.5 billion daily users across its platforms, generating substantial advertising revenue [6] - Meta's financial strength has allowed it to invest in growth initiatives, including the launch of a dividend in 2024 [7] Group 3 - Meta has recognized the potential of AI for growth and has increased its investment in this area, including building data centers and developing a large language model [10] - The establishment of Meta Superintelligence Labs, led by a notable talent from Scale AI, underscores the company's commitment to advancing its AI capabilities [10]

格隆汇1月21日｜Meta Platforms(META.US)首席技术官周三表示，该公司新成立的人工智能实验室本月 已向内部交付首批备受瞩目的人工智能模型。在达沃斯世界经济论坛年会间隙的简报会上，公司首席技 术官Andrew Bosworth称，这些由Meta超级智能实验室构建的模型"非常出色"。 ...

Core Insights - Traton Group reported a net cash flow of 1.643 billion euros ($1.92 billion) for 2025, which is a decline from the previous year but exceeds analyst consensus estimates of 1.011 billion euros [1] Financial Performance - The reported net cash flow of 1.643 billion euros represents a decrease compared to the prior year [1] - The performance surpassed analyst expectations, indicating stronger than anticipated financial health [1]

Core Insights - Meta Platforms has launched a new artificial intelligence lab that has successfully delivered its first high-profile AI models internally this month, as stated by the company's chief technology officer [1] Company Developments - The establishment of the new AI lab signifies Meta's commitment to advancing its capabilities in artificial intelligence [1] - The internal delivery of AI models indicates progress in Meta's AI initiatives and potential future applications [1]

格隆汇1月21日丨伟达涨1%，特斯拉涨0.8%，Meta、亚马逊涨0.4%，微软涨0.2%，苹果涨0.1%。 ...

Group 1 - The core viewpoint is that the AI glasses industry is entering a rapid growth phase, driven by technological breakthroughs and new product launches, leading to an upturn in the industry cycle for related companies [1][4] - The supply chain for AI glasses is more extensive than traditional eyewear, involving a larger number of participants, including hardware suppliers (chips, sensors, audio, Bluetooth), software providers, and ODM/OEM manufacturers [3] - XR startups typically collaborate with traditional eyewear manufacturers, with wearable device-related companies making up 45% of the entrants in the AI glasses market [3] Group 2 - The launch of Ray-Ban Meta smart glasses marks a new product cycle, featuring improved audio and video quality, and integration with Meta's social platforms [2] - Domestic AI glasses sales are rapidly increasing, with a reported 210,000 units sold in Q3 2025, a 200% quarter-on-quarter growth, driven by brands like Xiaomi and Ray-Ban Meta [4] - Xiaomi's AI glasses are projected to sell over 300,000 units by the end of 2025, while Alibaba's Quark AI glasses have seen strong pre-sale demand, with initial batches selling out quickly [4]

Investment Rating - The report maintains an "Outperform" rating for the AI industry [1] Core Insights - The capital expenditure (Capex) guidance from major overseas companies is optimistic, with a continuous upward trend in computing power expected through 2026. The total Capex for Microsoft, Google, Amazon, and Meta is projected to reach $406.5 billion in 2025 and $596.4 billion in 2026, representing year-on-year growth of 46% and 47% respectively [3][5][22] - The demand for AI-related infrastructure is driving significant upgrades in interconnects, cooling, and power supply sectors, indicating a long-term growth cycle distinct from previous technology cycles [3][4] Summary by Sections 1. Overseas Major Companies' Capex Guidance - Major companies like Microsoft, Google, Amazon, and Meta are expected to increase their Capex significantly, with projections of $186.4 billion, $139.5 billion, $162.5 billion, and $108 billion respectively for 2026 [6][7][14][18][21] - The overall Capex for these four companies is expected to account for 62% of the global AI-related Capex, which is projected to reach $960 billion by 2026 [22][23] 2. Interconnect Sector: Optical Modules and PCBs - The interconnect sector is experiencing a fundamental shift in demand due to AI server cluster construction, leading to simultaneous upgrades in computing boards, switches, and optical modules, which will increase both demand and pricing [3][4] - The deployment of 800G technology is expected to accelerate, with 1.6T technology entering the introduction phase [3] 3. Cooling Sector: Liquid Cooling Demand - The power density of AI GPU racks is projected to rise from 130 kW in 2024 to over 1 MW by 2029, making liquid cooling technology essential [3][4] 4. Power Supply Sector: AIDC Power Distribution Evolution - The evolution of AIDC power distribution methods is crucial, with a focus on HVDC and SST technologies to enhance system efficiency and reduce energy losses [3][4] 5. Investment Recommendations - The report suggests focusing on specific companies within various sectors: - Computing power: Haiguang Information - Interconnects - Optical modules: LightSpeed Technology, Huagong Technology - Interconnects - PCBs: Shenghong Technology, Shennan Circuits, and others - Cooling: Invec - Power supply: Magpow [3][4]

截至2026年1月21日11:08，电子ETF(159997)盘中换手2.14%，成交3044.34万元。跟踪的中证电子指数(930652)强势上涨3.21%， 成分股龙芯中科上涨20.00%，海光信息上涨13.80%，华天科技上涨10.01%，中国长城，通富微电等个股跟涨。 【产品亮点】 覆盖半导体+苹果产业链的电子ETF(159997)被动跟踪中证电子指数，重仓半导体、消费电子行业，汇聚AI芯片、汽车电子、 5G、云计算、印制电路板（PCB）等热门产业，前十大权重股囊括市场热议的工业富联、海光信息、胜宏科技，还包括"果 链"龙头立讯精密等。 此次扩产背后，是终端需求的爆发式增长与技术瓶颈的持续突破。Meta旗下Ray-BanMeta系列AI眼镜销量已突破200万副，2025 年上半年全球市占率高达73%，因库存紧张已暂停部分国际市场拓展。需求端，AI大模型落地推动设备向轻量化、智能化升 级，2025年中国AI眼镜销量同比增长215.7%，全球市场2025-2029年复合增长率预计超60%；技术端，MicroLED、衍射光波导 等核心技术成熟度提升，端侧AI算力突破实现本地大模型运行，大幅优化用户交互体验。 ...