Group 1 - Recent advancements in artificial intelligence (AI) include DeepSeek's introduction of "conditional memory" mechanisms, significantly enhancing the performance of large language models in knowledge retrieval and computational efficiency [1] - Alibaba's "Qianwen" app has achieved over 100 million monthly active users (MAU) within two months of launch, integrating with various Alibaba ecosystem services such as Taobao and Alipay [1] - Zhiyun Securities highlights that the acceleration of AI application commercialization is a core driver of increasing computing power demand, driven by the rapid evolution of domestic large model capabilities and the commercialization efforts of overseas giants like OpenAI [1] Group 2 - As of January 15, 2026, the CSI Sci-Tech Innovation and Entrepreneurship AI Index (932456) shows mixed performance among constituent stocks, with Hehe Information leading at a 3.45% increase [2] - The CSI Sci-Tech Innovation and Entrepreneurship AI ETF (588410) has a recent price of 1.13 yuan, with a turnover rate of 8.72% and a transaction volume of 38.094 million yuan during the trading session [2] - The top ten weighted stocks in the CSI Sci-Tech Innovation and Entrepreneurship AI Index, as of December 31, 2025, include companies like Zhongji Xuchuang and Xinyi Sheng, collectively accounting for 58.76% of the index [2]

Group 1: Market Overview - The global large model market is expected to reach $206.5 billion by 2029, with a CAGR of 80.7% from 2024 to 2029, driven by the iteration of large models and increased AI application penetration [1] - The large model application market is projected to be the main growth driver, with an estimated market size of $151.5 billion by 2029, reflecting a compound annual growth rate (CAGR) of 84.4% during the same period [1] Group 2: AI General Assistants - The AI general assistant sector is anticipated to see leading companies continue to consolidate the market, with OpenAI, Anthropic, Google, and xAI forming a first tier in overseas markets [2] - In the Chinese market, ByteDance, Tencent, Alibaba, and DeepSeek are expected to form a first tier, while smaller firms like Kimi and Minimax may face disadvantages in infrastructure and traffic acquisition as competition intensifies [2] - As of December 2025, active users for ByteDance's Doubao, DeepSeek, and Tencent's Yuanbao are projected to reach 155 million, 82 million, and 21 million respectively [2] Group 3: AI Entertainment Companions - The AI companionship applications are still in the early development stage, with rapid improvements in large model text and multimodal capabilities driving user penetration and consumption [3] - By 2030, the penetration rate of AI companionship applications among global internet users is expected to exceed 40%, with a projected annual revenue of approximately $70 billion, representing a CAGR of 200% from 2023 to 2030 [3] - Leading AI companionship applications include Character AI, Minimax Talkie AI, and ByteDance's Cat Box, with MAUs expected to reach 3.09 million, 2.35 million, and 470,000 respectively by December 2025 [3] Group 4: AI Visual Generation - The global AI visual generation application market is projected to reach $16.6 billion by 2027, with a CAGR of 66% from 2025 to 2027 [4] - Companies with leading model capabilities, product capabilities, and financial resources, such as Google Veo and Kuaishou's Keling, are expected to perform well in this sector [4] - Kuaishou's Keling AI is set to experience several catalysts, including the launch of new features and a revenue breakthrough of $20 million in December 2025, corresponding to an ARR of $240 million [4] Group 5: Minimax Valuation - Minimax completed its public offering in Hong Kong in January, increasing market attention on the AI application sector [5] - As of January 12, Minimax's valuation corresponds to approximately 222 times its estimated ARR for 2025 [5] - Kuaishou's Keling is conservatively estimated to have a valuation of $2.8 billion to $5.6 billion, representing about 6% to 18% of Kuaishou's market value, with a valuation discount of 70-90% compared to Minimax [5]

Core Insights - Artificial intelligence (AI) is a strategic technology driving a new wave of technological revolution and industrial transformation, significantly altering human production and lifestyle [1] - Shenzhen is positioning itself as a leading AI hub by creating an environment that supports comprehensive AI development, including supply chain efficiency, open application scenarios, and robust industrial ecosystems [1][3] Group 1: AI Ecosystem in Shenzhen - Shenzhen has over 2,600 AI companies, forming a tiered development structure led by major players like Tencent and Huawei, along with more than 10 AI unicorns and 34 listed robotics companies [3] - The city is focused on "chain supplementation" to address pain points in the transition from concept to product in the smart terminal sector, establishing a national AI application pilot base [3] - The "Moli Ying" AI ecosystem community in Nanshan is a leading incubator in China, providing comprehensive services to nearly 200 resident companies with a total valuation exceeding 20 billion [4] Group 2: Support for Startups - Shenzhen's policies are designed to support entrepreneurs with low-cost office spaces and a multi-tiered housing system, allowing talent to settle and thrive in the city [7] - The city has launched various funding initiatives, including a 500 billion yuan venture capital fund and specific grants for AI research and development, to facilitate the growth of startups [9] - Shenzhen's market cultivation system supports companies from inception to public listing, with an average time of 13.35 years for startups to become listed companies, which is faster than the national average [10] Group 3: Talent Development - Shenzhen is cultivating high-end, interdisciplinary talent through institutions like the Hong Kong University (Shenzhen) AI Academy, emphasizing practical training in design thinking and system engineering [8] - The city has attracted 220,000 returnees and 26,500 high-level talents, with a significant number of scholars recognized among the top 2% globally [8] Group 4: Innovation and Commercialization - Startups in Shenzhen are rapidly moving from technical prototypes to commercial markets, with notable success stories emerging from the "Moli Ying" community [5][6] - The city has established a mechanism for seamless integration of technology validation and commercialization, with nearly 300 high-value application scenarios available for enterprises [9]

截至2026年1月15日 13:44，中证人工智能主题指数(930713)成分股方面涨跌互现，合合信息领涨 3.14%，北京君正上涨2.32%，澜起科技上涨2.14%；中科星图领跌。AI人工智能ETF(512930)最新报价 2.38元。 数据显示，截至2025年12月31日，中证人工智能主题指数(930713)前十大权重股分别为中际旭创、新易 盛、寒武纪、澜起科技、中科曙光、科大讯飞、海康威视、豪威集团、金山办公、浪潮信息，前十大权 重股合计占比58.08%。 AI人工智能ETF(512930)，场外联接(平安中证人工智能主题ETF发起式联接A：023384；平安中证人工 智能主题ETF发起式联接C：023385；平安中证人工智能主题ETF发起式联接E：024610)。 风险提示：基金有风险，投资需谨慎。基金管理人承诺以诚实信用、勤勉尽责的原则管理和运用基金资 产，但不保证本基金一定盈利，也不保证最低收益。基金管理人提醒投资人基金投资的"买者自负"原 则，在做出投资决策后，基金运营状况与基金净值变化引致的投资风险，由投资人自行负担。基金的过 往业绩及其净值高低并不预示其未来业绩表现，基金管理人管理的其他基金 ...

本报记者 杨沙沙 编者的话：路透社近日报道称，中国人工智能（AI）初创公司深度求索（DeepSeek）预计将于2月中旬 推出其下一代人工智能模型V4，该模型具有强大的编码能力，表现可能优于美国人工智能公司 Anthropic旗下大模型Claude和美国初创公司OpenAI的GPT系列等竞争对手。一年前，DeepSeek发布大 模型R1，英国广播公司（BBC）形容"DeepSeek向世界展示了中国人工智能领域的竞争力"，当时距离 OpenAI推出ChatGPT也已两年。多名接受《环球时报》记者采访的专家、学者表示，2025年一年时间， 中国缩小了与美国在人工智能领域的差距，将一岁的DeepSeek和三岁的ChatGPT作为"时代的切片"，能 清晰看到中美站在不同的航道。 "中美大模型跑出两条主航道" 一年前，日本企业（中国）研究院执行院长陈言忽然发现，自己在中关村融科资讯中心办公大楼的"邻 居"——中国人工智能（AI）初创公司DeepSeek火了。电梯经过DeepSeek所在的楼层不再停靠，媒体记 者守在楼下等待采访，紧接着陈言收到大量日本企业想投资这家企业的请求。但陈言告诉《环球时报》 记者："这些日本企业已 ...

科大讯飞方面表示，未来，第三方开发者、行业专家可以基于此平台，开发并上架更丰富的垂直场景智 能体，共同打造招采领域的"智能应用商店"。 （文章来源：新京报） 近日，4名评标专家因为送餐延误导致身体不适的消息引发关注。官方通报此事时表示"评标工作责任 重、强度高，评标时长不确定"。值此当口，科大讯飞宣布推出"招采智能体平台"，在降低合规风险的 同时提升评标工作效率。 1月15日，科大讯飞方面告诉贝壳财经记者，招采业务场景千变万化，因此平台的终极目标是构建生 态，其内置了覆盖招标、投标、评标核心流程的数十个专业Agent。 在广受关注的评标领域，科大讯飞工作人员向记者展示了通过AI自动评审与推荐的流程。记者观察 到，AI可以自动对标书进行文本解析，数据提取和图像识别，并在文件评审流程进行多维度对比和AI 评分，最终可以输出一份智能评标报告，以及供应商优劣对比表和推荐的中标候选人（附理由）。 ...

15日早盘，AI题材集体回调，创业板人工智能早盘下探逾3%。其中，利空突袭，AI应用概念股纷纷大跌，8天翻倍大牛股易点天下停牌核查，易华录跌停， 蓝色光标、拓尔思大跌超10%。但算力方向局部活跃，兆龙互连、联特科技等多股涨超5%。 热门ETF方面，双线布局"算力+AI应用"的创业板人工智能ETF（159363）场内价格大跌3%，成交额超5亿元，资金继续进场抢筹，实时净申购1.4亿份，此 前3日已累计加码9.7亿元。 | | 多日 1分 5分 - | | | | | 19 益則溫后 肇加 九转 圖线 上具 (2) > | | | | | | 印度版人 智能ETP | | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 1.63 | | | 159363[创业板人工智能ETF华宝] 11:02 价 1.088 涨跌 | | | -0.036(-3.20%) | | 1.088 | -0.036 -3.20% SZSE CNY 11:02:12 交易中 查看L2全导 | | | | 159 ...

前段时间 Meta 要 砸下二 十亿刀收购 AI Agent 公司 Manus 后，虽然现在出了点问题，还没尘埃落定，但是 Manus 的价值让太多人眼馋了～ Manus 自 2025 年 3 月正式发布以来，仅 八个月就突破约 1 亿美元 ARR（年经常性收入） ，并积累了数百万用户，该公司开发的 AI 代理能够 自动完成从 市场调研、数据分析到编码等复杂任务，而非仅仅进行对话式应答。 有人总结了 Manus 的发展逻辑，还挺有道理： 这操作 666～ 一个产品火出圈 后，技术圈 总会 生出一个开源版～ 比如 planning-with-files 这个开源项目就复刻了 Manus 的核心工作流，上线没 几 天就收获 7.5k+ 的 Star Markdown 是我在硬盘上的工作记忆载体。信息处理本身是一个迭代推进的过程，而实时上下文窗口的承载能力有限，因此 Markdown 文件既是随手记录的草稿本、追踪进度的检查点，更是搭建最终交付成果的核心基石。 关注度持续提升： planning-with-files 是一款 Claude Code 插件 ，它能重构你的工作流，通过 持久化 Markdown 文件 ...

15日早盘，AI题材集体回调，创业板人工智能早盘下探逾3%。其中，利空突袭，AI应用概念股纷纷大 跌，8天翻倍大牛股易点天下停牌核查，易华录跌停，蓝色光标、拓尔思大跌超10%。但算力方向局部 活跃，兆龙互连、联特科技等多股涨超5%。 热门ETF方面，双线布局"算力+AI应用"的创业板人工智能ETF（159363）场内价格大跌3%，成交额超5 亿元，资金继续进场抢筹，实时净申购1.4亿份，此前3日已累计加码9.7亿元。 立足当下，创业板人工智能板块仍有多重催化：AI应用加速商业化，有望持续兑现；算力板块同步迎 来价值重估：IDC受益于需求复苏与估值修复，光模块产能释放驱动业绩增长。AI与算力产业正进入协 同发展新阶段，创业板人工智能配置价值凸显。 当前AI产业动态密集，海外xAI、Anthropic陆续融资，国内"人工智能+制造政策"推出，智谱和MiniMax 上市大涨，后续DeepSeek-V4推出有望引发新一轮AI应用热潮。中信建投表示，随着模型能力不断提 升，尤其是推理与长窗口成本的显著下降，AI下游应用场景正加速进入商业化验证阶段。* IDC（数据中心）这一低位算力板块明显异动，有望迎重新定价。国盛证券 ...

智通财经APP获悉，智谱(02513)逆市涨超6%，截至发稿，涨6.39%，报229.8港元，成交额3.35亿港元。 东吴证券认为，作为纯大模型玩家，公司受益于云端规模效应和Agent/编程场景红利。看好智谱AI在本 土大模型技术实力、开源生态布局以及政企本地化落地能力方面的优势，并认为公司有望受益于中国大 模型行业从本地化部署向云端服务转型的长期趋势。 消息面上，智谱日前宣布联合华为开源新一代图像生成模型GLM-Image，模型基于昇腾Atlas 800T A2设 备和昇思MindSpore AI框架完成从数据到训练的全流程，是首个在国产芯片上完成全程训练的SOTA多 模态模型。据介绍，GLM-Image实现了图像生成与语言模型的联合，API调用模式下，生成一张图片仅 需0.1元。 ...