" 有数据显示，朴朴超市估值已达到60亿元+，相当于两个叮咚买 菜的市值。 " 作者丨代聪飞 编辑丨林觉民 一位熟悉朴朴的人士表示，在朴朴起家的福建，朴朴超市的业务数据很早就可以达到外卖业务一半，由此 对整个即时零售行业有了相当积极的判断，认为美团的闪购未来可以达到一个小外卖的体量。 根据公开信息，朴朴超市自成立来完成五轮融资，最近的一轮也发生于2021年。但据雷峰网了解，朴朴超 市近两年在资本领域动作并不少，除却没达成的收购，朴朴在去年也完成了新一轮融资。 与阿里的盒马，美团的小象超市，叮咚买菜等相比，朴朴的扩城速度可谓缓慢，迄今为止仅覆盖福州、厦 门、泉州、广州、佛山、深圳、东莞、武汉、成都等九大城市，其中广深收入更是占比高达70%，2025 年仅新开泉州一城。 扩张缓慢的一个重要原因，或许与朴朴超市坚持的大前置仓和区域运营模式相关，但这也是朴朴超市能够 壮大发展至今的关键。 朴朴超市的前置仓采用的是1000平左右的大仓库，实现6000-8000的SKU。这样的SKU数量已经超过了 传统中大型超市。相较主打生鲜的小仓，朴朴可以实现更高的客单价。在福州，朴朴的客单价可以达到人 均80。 近日，有消息称，朴朴 ...

" 地主家也没余粮了。 " 作者丨覃倩雯 编辑丨刘伟 "地主家也没余粮了。"接近亚马逊人士向雷峰网透露，整个HC确实锁了不少，大中华地区的招聘需求表 从以前的二十多页，缩减到现在的两三页，划两下就没了。 近日，社交平台上不少信息显示，亚马逊中国内部正在陆续裁员，涉及ESM、LCS等部门。据雷峰网独家 获悉，目前亚马逊中国已经暂停了大部分的招聘需求，各部门的人员预算也砍了很多。 "我的经理上个月刚被裁，职级是L7，但直接下属只有6个人，被裁得很突然。"一位亚马逊员工说道。另 一位亚马逊员工透露，我们的L7是去年12月走的，跑得很快，不想转组变庶子。 目前，亚马逊中国的核心零售业务板块只有广告团队和招商团队，以销售功能为主；而海外的美国总部的 团队是负责整个亚美的运营、采销、流量分发等。 上述知情人士透露，其实亚马逊的政策中国区这帮人也没办法左右，中国区现在很多权限都没了，决策权 都在国外，他们大部分都是做一些执行的东西，很多人的业绩目标都是海外那边的负责人定的。（欢迎添 加作者微信 qqw_501， 一起交流Temu、亚马逊内部更多信息。） 组织调整之下，亚马逊与Temu短兵相接 巨变背后，是亚马逊零售业务发展 ...

Group 1 - The automotive industry is currently facing challenges, with dealers struggling to profit from selling new energy vehicles, leading to a reduction in players in the market [4][5] - Zero Run's strategy focuses on cost reduction and efficiency improvement, with a shift from traditional inventory models to order-based inventory management [4][5] - The high-end positioning of Chinese car brands is hindered by a lack of cultural identity and brand value, as stated by Great Wall Motors' chairman [9][10] Group 2 - Qualcomm announced its entry into the data center market, emphasizing its ongoing partnership with Xiaomi for flagship devices [11][12] - Xiaomi's upcoming product, the Xiaomi 15S Pro, is expected to incur significant R&D costs without recovering them, as it features extensive upgrades [13] - Hisense has appointed a former executive from Midea to strengthen its air conditioning business, indicating a strategic move to enhance competitiveness [14] Group 3 - NIO's founder addressed rumors about his death, emphasizing the challenges of dispelling false information in the entrepreneurial landscape [19] - CATL's recent IPO on the Hong Kong Stock Exchange saw its stock price rise by 12.55%, reflecting strong market interest [18] - A major restructuring at FAW-Volkswagen aims to enhance user service and product marketing capabilities in preparation for the launch of next-generation electric vehicles [16][17] Group 4 - Intel's CEO shared insights on cultural differences in business practices during a recent event in Taiwan, highlighting the importance of local customs [24] - NVIDIA's CEO criticized U.S. restrictions on AI technology, arguing that they could hinder the country's competitive edge in the field [25] - Foxconn plans to invest $1.5 billion in India to expand its manufacturing capabilities, particularly for iPhone assembly [27][28] Group 5 - OpenAI faces challenges in attracting investors due to its non-profit structure, which complicates its funding efforts [28] - Apple is internally testing its AI chatbot, aiming to compete with ChatGPT, with significant advancements reported [29] - Intel is considering divesting its Network and Edge Computing business to focus on PC and data center chips, reflecting a strategic shift [30]

Core Viewpoint - The article discusses the launch of Tencent's AI-driven game visual generation platform, "Hunyuan Game," which aims to industrialize game art production and improve efficiency in the gaming industry [2][4]. Group 1: Industry Context - The current game art production process involves multiple stages such as concept art, modeling, and scene creation, often requiring collaboration among different roles and tools [2]. - For small and independent developers, producing art assets is challenging, often relying on unstable AI models or outsourcing, leading to unpredictable costs [2][4]. Group 2: Hunyuan Game Features - Hunyuan Game integrates various stages of art production, allowing designers to input text or sketches and complete the entire workflow from concept exploration to 3D modeling without switching tools [2]. - The platform can generate character concept art and 3D models based on simple prompts, significantly streamlining the production process [2][5]. Group 3: Current Limitations - Despite its capabilities, AI-generated art still suffers from inaccuracies, requiring significant manual adjustments to meet artistic standards [4]. - Many established game studios continue to rely heavily on human artists for finalizing designs, indicating that AI's role is currently limited in the detailed refinement stages [4][5]. Group 4: Potential Impact - Hunyuan Game is expected to enhance efficiency in producing standardized art assets for large-scale AAA games, where the predictability of output is higher [5]. - Independent developers may benefit from faster prototyping and iteration at lower costs, although the emphasis on unique artistic styles may lead them to seek alternative solutions [6].

Core Viewpoint - The article discusses the challenges and advancements in the Mixed Expert Model (MoE) technology, particularly focusing on the load balancing issues and the introduction of the OmniPlacement strategy by Huawei to enhance inference performance [2][4][12]. Group 1: Challenges in MoE Models - The MoE models face significant challenges, particularly the "cold and hot expert" phenomenon, where some experts are frequently called (hot experts) while others are rarely used (cold experts), leading to uneven load distribution [2][4]. - This imbalance results in increased inference latency and limited throughput, as underutilized resources restrict overall system performance [3][14]. Group 2: OmniPlacement Strategy - Huawei's OmniPlacement strategy addresses these challenges through expert reallocation, inter-layer redundancy deployment, and near-real-time dynamic scheduling, significantly improving MoE model inference performance [4][12]. - The strategy includes a joint optimization algorithm that reduces load imbalance by analyzing expert activation data and optimizing deployment order based on call frequency and computational needs [5][14]. Group 3: Key Features of OmniPlacement - OmniPlacement employs inter-layer redundancy deployment to alleviate the pressure on hot experts by allocating additional redundant instances, thus enhancing system throughput [5][12]. - The framework supports dynamic resource allocation based on real-time resource usage and expert call frequency, allowing for predictive resource distribution to minimize performance discrepancies between hot and cold experts [6][9]. Group 4: Testing and Results - Comprehensive testing on the DeepSeek-V3 model demonstrated that OmniPlacement reduces average inference latency by approximately 10% compared to baseline methods, primarily due to dynamic expert allocation and communication domain optimization [12][14]. - The system's throughput improved by about 10%, reflecting a significant increase in resource utilization, especially in high-concurrency scenarios [14]. Group 5: Future Directions - Future research will focus on developing smarter scheduling algorithms and adaptive expert selection mechanisms to further enhance the system's adaptability to complex inputs [15][16]. - The OmniPlacement framework aims to expand its functionality to support more types of MoE models, increasing its versatility and applicability in various industrial settings [16].

Key Points - CATL prefers institutional investors over retail investors for its IPO, limiting retail participation to 7.5% despite high demand [4][5][6] - Xiaomi has invested over 13.5 billion RMB in its semiconductor project, aiming for top-tier performance with its new chip, the Xiaomi Xuanjie O1, which utilizes 3nm technology [8][9] - Huawei launched two HarmonyOS computers, including a foldable model starting at 23,999 RMB, marking a significant step in its ecosystem development [11][12][14] - Ant Group's international business, Ant International, generated nearly 3 billion USD in revenue last year and is preparing for a potential IPO [14][15] - Xiaomi reported a criminal gang manipulating nearly 10,000 accounts to defame the company, leading to multiple arrests [19][20] - Tesla's autonomous taxi service in Austin will operate on an invitation-only basis with a limited fleet of 10 to 20 vehicles [27] - Google I/O is set to showcase new AI models and updates to Android 16, emphasizing the integration of AI in search and other applications [25][26]

" 华为不只是「官宣」一下而已，后面更会是全面开源。 " 作者丨李希 推理部署，成为大模型落地重中之重 从 2017 年 Google 提出 Transformer —— 这一人工智能 中最常用的 神经网络架构，到 DeepSeek V3/R1 在 202 5 年 春节 一夜爆火，超大规模 MoE 架构大模型的重点逐渐从训练开发转向推理支撑的 应用落地。 推理场景是大模型认知能力的 " 试金石 " ，是大模型商业化落地的核心能力，从抢先上线 DeepSeek 模 型到 API 服务价格战，在推理为王的时代，谁能最极致的提升推理部署计算效率，谁才能真正获得大模型 商业成功。 数学补物理，极致提升计算效率 " 数学补物理 " ，通常指通过数学理论、算法和建模方法，弥补传统物理设备开发在复杂系统分析、大规 模计算或多场耦合问题中的局限性。华为轮值董事长孟晚舟曾在 2025 年新年致辞中提到： " 华为十多个实验室与伙伴们的工程师组成 " 大杂烩 " 团队，面对天成 AI 集群系统和单芯片性能的严峻 工程挑战，他们创造性应用数学补物理、非摩尔补摩尔、系统补单点等思想，在散热、供电、高速、高密 及大芯片在板可靠性等工程 ...

Core Viewpoint - Alibaba's AI division has maintained triple-digit growth for seven consecutive quarters, yet the secondary market remains unsatisfied, as evidenced by an 8% drop in stock price following the Q4 FY2025 earnings report [2][9]. Summary by Sections Financial Performance - Alibaba reported a 7% year-on-year increase in overall revenue, with Alibaba Cloud revenue reaching 30.127 billion yuan, marking an 18% growth, the fastest in three years, driven by increased adoption of AI-related products [2]. - For the entire fiscal year, Alibaba Cloud's revenue reached 118 billion yuan, achieving double-digit growth [2]. Investment in AI and Cloud Infrastructure - Alibaba plans to invest over 380 billion yuan in cloud and AI hardware infrastructure over the next three years, indicating strong commitment to these sectors [4]. - The company has been actively seeking partners for infrastructure development and has launched several AI models, including Qwen2.5-Max and Qwen3 series, to meet growing demand [4]. Market Dynamics and Customer Base - Alibaba Cloud's AI market strategy focuses on enterprise applications, expanding from large enterprises to small and medium-sized businesses across various industries, including traditional sectors [6]. - The company has been recruiting executives from major cloud providers like AWS and Microsoft to enhance its capabilities in the B2B market [6]. Market Expectations and Stock Performance - Despite strong performance metrics, the market's expectations for Alibaba Cloud's growth were deemed overly optimistic, leading to a significant stock price decline [9]. - Analysts noted that the capital expenditure for the quarter was 24.6 billion yuan, lower than Tencent's 27 billion yuan, raising concerns about Alibaba's enthusiasm for AI investments [9]. Profitability and Competitive Landscape - The profitability of AI products remains uncertain, as the revenue base is low and the return cycle for B2B clients is lengthy [9]. - Price reductions for AI services, such as a 97% decrease in the cost of Qwen-Long model inputs, have raised concerns about profitability in a competitive landscape [10]. - Despite challenges, Alibaba Cloud has managed to achieve profitability while other domestic cloud providers continue to incur losses [10]. Long-term Outlook - The pursuit of AI infrastructure in the B2B sector is characterized by high investment, long cycles, and uncertain returns, suggesting that the recent stock price drop may reflect a market correction regarding AI profitability expectations [11].

「 RTK+双目 」成割草机方案香饽饽，公司A和B上演 「 始乱终弃 」 九号割草机出了一个 「 RTK+双目」的方案，A公司认为这个方案可行，直接在内部优化了这个方案，并 投入了200人，甚至还找了外部协作团队B一起做这个事情。但等B把方案做出来，A觉得这个方案自己也 能做，加上B的工作效率不如预期，开价100万价格太高，就又自己做了。 据说公司A没有用B代工的还有一个顾虑是，B用了一家大厂的算法，担心后续会在法庭相见。（B公司长 于算法，劣于渠道，在做方案还是做品牌，做何种方案之间均有摇摆，A公司今年承压，尤其在九号侵入 其腹地之后，更多行业动态可添加微信 MOON_ERS 交流） 扫地机公司C一级代理商吃肉，二级代理商被吃干抹净 扫地机公司C采取 「一省一代理 」的渠道策略，C给代理商是30个点，达成后月返季返，最多能拿到32 个点。C的一级代理如果要拓展市场，会找二级代理商，但最多只能给22个点，如果按照国补15个点去 卖，合起来相当于30个点。所以如果没有国补，C的产品基本卖不动了。有人评价，C的一级代理商吃肉 喝血，二级代理商则是直接被吃干抹净。（今年扫地机市场的代理体系暗流涌动，C公司重投线上， ...

Core Viewpoint - The article discusses the shift in the gaming industry from content-driven products to platform-driven products, highlighting the challenges faced by major game developers in launching new titles that replicate past successes [2][5]. Group 1: Industry Overview - In April 2025, 33 mobile game publishers generated $2 billion, capturing nearly 40% of global publisher revenue [5]. - The current market is characterized by a "mobile gaming small year," with new heavy mobile games failing to perform as expected due to a lack of innovative content [5][6]. Group 2: Company Strategies - Major companies like Tencent are focusing on long-term operations of flagship products, such as "Honor of Kings," which saw a 71% increase in global revenue in April [6]. - NetEase is emphasizing self-developed games and international expansion, as seen with "Marvel Duel," which topped the Steam global sales chart after seasonal updates [7]. - MiHoYo maintains its position in the rankings through flagship projects like "Genshin Impact," which significantly boosted overseas revenue by nearly 200% [7]. Group 3: Market Dynamics - The gaming industry is witnessing a trend where casual and hyper-casual games are gaining traction, attracting a broader user base beyond traditional gamers [8]. - The competition in the overseas market is intense, with higher performance expectations for products compared to the domestic market [9]. - Smaller game developers are finding success with light games, although the profit margins and return on investment timelines are challenging for larger companies [9].