Core Viewpoint - The mobile phone manufacturers are shifting their focus towards the professional camera and handheld imaging device market, which was previously considered untouchable, indicating a significant transformation in the industry landscape [1]. Group 1: Market Dynamics - OPPO and vivo are the first to target the market dominated by DJI's Pocket series, which has become a standard for content creators and tourists seeking better image quality and stabilization than smartphones [2][4]. - vivo is developing a standalone Vlog camera aimed at competing directly with DJI's Pocket series, featuring a lightweight design and integration with its OriginOS for seamless editing and sharing [4][6]. - OPPO is also working on a new imaging product that may be part of its Find series, focusing on high-quality material collection while leveraging mobile processing power for editing and social sharing [8]. Group 2: Product Innovations - Xiaomi is pursuing a modular approach with a magnetic lens system that allows its smartphones to transform into high-quality cameras, potentially launching with the MIX 5 or 16 Ultra in 2026 [9][12]. - The magnetic lens module will enable users to attach a large optical module to their phones, enhancing photography capabilities while maintaining a lightweight design [13][15]. - Honor is taking a more radical approach by integrating a mechanical gimbal into its upcoming "Robot Phone," allowing for advanced stabilization and 360-degree tracking [17][19]. Group 3: Strategic Collaborations - Huawei is not developing a standalone Vlog camera but is collaborating with DJI for deeper integration of camera controls within its HarmonyOS, indicating a strategic focus on software rather than hardware [21][26]. - Huawei is also testing a square sensor for front cameras, aimed at optimizing social media content creation by allowing for maximum image capture regardless of phone orientation [23][24]. Group 4: Industry Trends - The year 2026 is seen as a pivotal moment for the smartphone industry, as manufacturers face diminishing returns on hardware improvements and are compelled to innovate through external devices or standalone products [27]. - The shift towards integrating advanced computational photography into traditional camera forms represents a significant challenge to established camera manufacturers, who have struggled to meet the demands of modern content creators [28][29].

Investment Rating - The report does not explicitly state an investment rating for the industry or specific companies involved in the research Core Insights - The Engram model proposed by DeepSeek and Peking University introduces a "Conditional Memory" mechanism that separates static knowledge recall from complex computations, significantly improving computational efficiency and task performance [1][2] - Engram-27B demonstrates systematic improvements over MoE-27B across multiple benchmarks, particularly excelling in long-context tasks [1][3] - The architecture allows for the offloading of large parameter tables to host memory, maintaining controllable inference throughput impact, thus validating the feasibility of "separation of storage and computation" [1][6] Summary by Sections Event - In January 2026, DeepSeek and Peking University released a paper on the Engram model, achieving significant performance improvements in various benchmarks while maintaining computational efficiency [1][17] Commentary - Engram innovatively decouples the recall of fixed knowledge from complex model computations, allowing models to focus on deeper reasoning tasks, thus enhancing overall efficiency [2][18] Performance Optimization - The study reveals an optimization path for resource allocation, indicating that transferring some model capacity to a conditional memory module can lead to a "U-shaped" performance trend, with a clear optimal performance range [3][19] - Replacing approximately 20% of traditional parameter capacity with conditional memory can yield significant improvements in knowledge-intensive tasks [3][19] Long Context Processing - Engram effectively offloads local repetitive details to memory lookup, allowing the backbone network to focus on global information integration, which is crucial for long-text processing [4][20] - In experiments, Engram-27B consumed only about 82% of the baseline pre-training computation while achieving higher accuracy in long-text retrieval tasks [4][20] System-Level Design - Engram's deterministic addressing mechanism allows for data pre-fetching from host memory, alleviating pressure on high-bandwidth memory (HBM) and controlling inference overhead to within 3% even with large memory tables [6][22] - The innovation shifts the focus from GPU memory constraints to CPU memory capacity and interconnect technologies, potentially redefining the critical constraints of AI systems [6][23] Impact on Chinese Large Models - Engram's ability to transfer memory-type parameters to scalable system memory enhances model capabilities while reducing reliance on high-end HBM, providing a clearer path for efficiency-driven technological advancement in China's large model industry [7][24] - The open-sourcing of related papers and code lowers barriers for industry validation and development, facilitating faster deployment and commercialization of large models in cost-sensitive environments [7][26]

Investment Rating - The report maintains a "Hold" rating for the CPU industry, indicating a cautious outlook on the sector's performance relative to the benchmark index [8]. Core Insights - The CPU's importance is expected to significantly increase due to advancements in AI and the emergence of new computational scenarios, which demand higher processing speeds and precision [13]. - DeepSeek's recent developments in AI model architecture highlight the potential for CPUs to handle large-scale models more efficiently, reducing reliance on expensive GPU memory [5][6]. - The ongoing shortage of CPU supply, particularly from leading manufacturers like Intel, is projected to peak in the first quarter of 2026, driven by demand exceeding supply [6][7]. Summary by Sections CPU Demand and Supply Dynamics - The demand for CPUs is anticipated to rise due to the increasing need for processing power in AI applications, with Intel indicating that the shortage may persist as they do not plan to expand production capacity [6][7]. - The report notes that the design of systems that decouple storage and computation can lead to significant efficiency gains, allowing for larger model parameters to be stored in cost-effective CPU memory [5][6]. AI Agent Trends - The report discusses the expected growth in the number of active AI agents, predicting an increase from approximately 28.6 million in 2025 to 2.216 billion by 2030, which will drive CPU demand [12]. - The annual execution of tasks by these agents is projected to rise dramatically, necessitating a substantial increase in CPU supply to meet the demands of AI workloads [12]. Investment Recommendations - The report suggests focusing on several key areas within the CPU ecosystem, including CPU chip manufacturers, advanced wafer fabrication, and related solutions, highlighting companies such as Haiguang Information and SMIC [13].

Core Viewpoint - The article discusses a groundbreaking paper by DeepSeek and Peking University that introduces a new module called Engram, which separates memory from computation in AI models, leading to a significant increase in reasoning capabilities [3][12]. Group 1: Introduction of Engram Module - DeepSeek's Engram module represents a supply-side reform in AI model architecture, allowing static knowledge to be stored separately from computational tasks, thus enhancing AI's reasoning abilities [3][14]. - The Engram module is inspired by the classic N-gram concept from natural language processing, modernized to allow for efficient retrieval of static knowledge with a time complexity of O(1) [15][16]. Group 2: Technical Innovations - Engram utilizes a large, scalable embedding table to store static knowledge, allowing for direct retrieval without complex computations, contrasting with traditional Transformer models where knowledge is embedded in weights [18]. - Three technical barriers were addressed: - A. Vocabulary compression reduced the effective vocabulary size by 23% through normalization of semantically similar terms [19]. - B. Multi-head hashing resolves hash collisions by mapping multiple N-grams to limited memory slots, enhancing robustness [20]. - C. Context-aware gating acts as a referee to filter out irrelevant static knowledge based on the current context [21][22]. Group 3: Resource Allocation and Model Performance - A large-scale ablation study revealed a U-shaped scaling law for resource allocation, indicating that the optimal distribution of parameters is approximately 75%-80% for Engram and 20%-25% for MoE, minimizing loss [30][31]. - The introduction of Engram not only improved knowledge tasks but also unexpectedly enhanced performance in logic, coding, and mathematics, with significant score increases across various benchmarks [39][40]. Group 4: Engineering Breakthroughs - Engram's architecture allows for a separation of memory and computation, enabling large models to offload memory to cheaper, scalable CPU resources, thus reducing reliance on expensive GPU memory [46][49]. - This separation allows for prefetching of memory data, maintaining high throughput even with large parameter sizes, which is a significant advantage for future AI model development [51][52]. Group 5: Future Implications - The upcoming DeepSeek V4 model is expected to integrate Engram technology, achieving a balance between computation and memory, enhancing both knowledge capacity and reasoning capabilities while reducing inference costs [61][64]. - The paper signals a shift in the AI industry towards architectural innovation, moving away from merely increasing computational power and parameters, and redefining competitive standards in AI development [65].

Core Viewpoint - Fujian Province is advancing measures to develop computing power infrastructure, focusing on enhancing data storage capabilities and promoting innovative technologies in data management [1] Group 1: Data Storage Enhancements - The initiative emphasizes strengthening data storage capacity and conducting maturity research and evaluation for data centers [1] - It aims to promote cutting-edge technologies such as all-flash and Blu-ray storage, optimizing algorithms for data compression and encryption to reduce storage usage while improving efficiency and security [1] - By 2027, the target is to achieve over 35% of advanced storage capacity [1] Group 2: Infrastructure Development - The plan includes the formulation of standards for classifying and grading hot, warm, and cold data, as well as specifications for tiered storage [1] - There is a focus on the domestic upgrade and transformation of core computing and storage network facilities [1] - Key cities such as Fuzhou (including Pingtan), Xiamen, and Quanzhou will explore a metropolitan "separation of storage and computing" model over a hundred kilometers [1]

Core Insights - The financial industry is undergoing a profound paradigm shift driven by real-time trading, intelligent risk control, and personalized services, which are now essential capabilities rather than mere concepts [1] - The "2025 Smart Finance Core Database Transformation Practice Seminar" held in Suzhou gathered industry experts to explore the latest trends and pathways for core system transformation [1] - There is a consensus in the industry that the separation of storage and computing has evolved from a technical option to a strategic cornerstone, defining the pace and pattern of financial innovation for the next decade [1] Group 1: Challenges of Traditional Systems - The traditional tightly-coupled architecture in the financial industry has reached its limits, necessitating a break from the "storage-computing integration" to fully unleash data potential [2] - Key issues with the existing architecture include performance bottlenecks, lack of agility, and operational complexity, which threaten business continuity [3] Group 2: Advantages of Storage-Computing Separation - Storage-computing separation is not merely a technical fix but a systemic upgrade that enhances data flow and system responsiveness [4] - This architecture provides four core values: extreme reliability, flexible elastic scaling, simplified operational management, and an open ecosystem that supports various database types [4][5] Group 3: Evolution of Databases - The new generation of databases, when combined with storage-computing separation, can significantly enhance performance, as demonstrated by a rural commercial bank that increased its transaction capacity from 1 million to 15 million daily transactions [6] - Modern databases support mixed workloads and are evolving towards advanced distributed architectures, showcasing substantial improvements in transaction processing capabilities [6] Group 4: Future of Financial Technology - The financial industry is witnessing a structural migration from traditional systems to intelligent frameworks, marking a transition from "accounting" to "intelligence" [8] - As of the end of 2024, distributed database instances in Chinese banks have increased from 3.9% in 2018 to 23.8%, with some banks reaching 100%, indicating a shift towards a new phase centered on intelligent infrastructure [8] - The ongoing digital transformation in finance is not just about cloud systems and intelligent applications but also about reconstructing the foundational architecture for the next decade [8][9]

Core Insights - Suzhou Rural Commercial Bank, established in 2004, has evolved into a key player in local financial services, focusing on rural finance and innovation through technology [1][3] - The bank is undergoing a digital transformation to overcome challenges posed by traditional banking structures and increasing competition from larger banks and internet financial services [3][5] Group 1: Digital Transformation Challenges - The bank's traditional systems are becoming bottlenecks, hindering business growth and innovation due to outdated infrastructure [3][5] - There is a pressing need for a more flexible and efficient core banking system to meet the demands of online marketing and digital risk control [3][5] Group 2: Technological Collaboration - Suzhou Rural Commercial Bank has partnered with Huawei to implement a "separation of storage and computing" architecture, utilizing GaussDB and OceanStor Dorado to enhance system performance and security [5][6] - This new architecture allows for centralized management of storage resources, significantly improving system elasticity and reducing total cost of ownership (TCO) [5][6] Group 3: Performance Improvements - The implementation of the new system has led to a transformation in the bank's IT infrastructure, enabling faster loan approvals and seamless transaction processing [6][7] - The bank can now offer real-time, accurate financial services, enhancing customer experience and operational efficiency [7] Group 4: Strategic Positioning - The collaboration with Huawei positions Suzhou Rural Commercial Bank as a national benchmark for financial transformation, showcasing how smaller banks can leverage technology for growth [6][7] - The bank's focus on innovation and digitalization is expected to drive its future success and strengthen its competitive edge in the financial sector [6][7]

Core Insights - Suzhou Rural Commercial Bank, in collaboration with Huawei, has completed a significant transformation of its core system using Huawei's GaussDB database and OceanStor Dorado all-flash storage, marking a key step for regional banks in China [1][3] - The core database is identified as the "heart" of financial IT systems, essential for digital transformation, with a focus on self-innovation, high availability, and performance [1][3] - The bank's core system has achieved a smooth transition from traditional commercial databases to GaussDB, supporting daily transaction volumes of tens of millions, with system performance improved by over 40% and recovery times reduced to seconds [5][7] Industry Challenges and Solutions - The financial industry faces three main challenges: accelerated business iteration requiring higher system elasticity, frequent security threats necessitating stronger architecture, and regulatory and cost pressures driving efficient infrastructure evolution [3] - Huawei emphasizes the need for a self-innovative core database system to maintain development control, proposing the "GaussDB + separation of storage and computing" architecture as a solution [3] Practical Implementation and Recommendations - A white paper released during the event outlines the technology selection, deployment models, and risk control points for regional banks undergoing transformation, recommending a "gradual replacement + dual-active disaster recovery" strategy [5] - Jiang Wei, Assistant General Manager of the Financial Technology Department at Suzhou Rural Commercial Bank, highlighted the successful implementation of a dual-active deployment model, which enhances system resilience and operational efficiency [5][7] Broader Implications - The experience shared by Suzhou Rural Commercial Bank serves as a model for small and medium-sized banks, balancing security, performance, and cost-effectiveness in their transformation efforts [7] - The collaboration between Huawei and Suzhou Rural Commercial Bank is expected to provide a valuable reference path for regional banks in their self-research transformation and digital competitiveness [7]

Core Concept - The article discusses the concept of "Compute In Memory" (CIM), which integrates storage and computation to enhance data processing efficiency and reduce energy consumption [1][20]. Group 1: Background and Need for CIM - Traditional computing architecture, known as the von Neumann architecture, separates storage and computation, leading to inefficiencies as data transfer speeds cannot keep up with processing speeds [2][10]. - The explosion of data in the internet era and the rise of AI have highlighted the limitations of this architecture, resulting in the emergence of the "memory wall" and "power wall" challenges [11][12]. - The "memory wall" refers to the inadequate data transfer speeds between storage and processors, while the "power wall" indicates high energy consumption during data transfer [13][16]. Group 2: Development of CIM - Research on CIM dates back to 1969, but significant advancements have only occurred in the 21st century due to improvements in chip and semiconductor technologies [23][26]. - Notable developments include the use of memristors for logic functions and the construction of CIM architectures for deep learning, which can achieve significant reductions in power consumption and increases in speed [27][28]. - The recent surge in AI demands has accelerated the development of CIM technologies, with numerous startups entering the field alongside established chip manufacturers [30][31]. Group 3: Technical Classification of CIM - CIM is categorized into three types based on the proximity of storage and computation: Processing Near Memory (PNM), Processing In Memory (PIM), and Computing In Memory (CIM) [34][35]. - PNM involves integrating storage and computation units to enhance data transfer efficiency, while PIM integrates computation capabilities directly into memory chips [36][40]. - CIM represents the true integration of storage and computation, eliminating the distinction between the two and allowing for efficient data processing directly within storage units [43][46]. Group 4: Applications of CIM - CIM is particularly suited for AI-related computations, including natural language processing and intelligent decision-making, where efficiency and energy consumption are critical [61][62]. - It also has potential applications in AIoT products and high-performance cloud computing scenarios, where traditional architectures struggle to meet diverse computational needs [63][66]. Group 5: Market Potential and Challenges - The global CIM technology market is projected to reach $30.63 billion by 2029, with a compound annual growth rate (CAGR) of 154.7% [79]. - Despite its potential, CIM faces technical challenges related to semiconductor processes and the establishment of a supportive ecosystem for design and testing tools [70][72]. - Market challenges include competition with traditional architectures and the need for cost-effective solutions that meet user demands [74][76].

Market Overview - On August 1, the market experienced fluctuations with the three major indices slightly declining, and the total trading volume in the Shanghai and Shenzhen markets was 1.60 trillion CNY, a decrease of 337.7 billion CNY compared to the previous trading day[2] - The Shanghai Composite Index saw a net outflow of 2.381 billion CNY, while the Shenzhen Composite Index had a net inflow of 2.675 billion CNY[4] Sector Performance - The sectors with the highest net inflows were photovoltaic equipment, traditional Chinese medicine, and logistics, while the sectors with the highest net outflows included semiconductors, components, and ground weaponry[4] - The pharmaceutical and education sectors showed resistance but did not exhibit complete trends, indicating potential areas for continued observation[1] Economic Policy Insights - The National Development and Reform Commission (NDRC) announced that the 800 billion CNY list of "two heavy" construction projects has been fully allocated, and 735 billion CNY of central budget investment has been largely distributed[5] - The NDRC plans to implement a "AI+" initiative to enhance the application of artificial intelligence, indicating a focus on technological advancement[5] Long-term Investment Directions - Long-term investment opportunities are expected to be centered around industries supported by national policies, particularly in new energy and computing power sectors[1] - The NDRC is also working on establishing a list of national-level zero-carbon parks, which may present future investment opportunities[5] Fund Dynamics - The second batch of floating fee funds is set to launch, with three products scheduled for issuance on August 4, including a medical innovation fund with a fundraising cap of 3 billion CNY[12] - The number of private equity securities investment funds from insurance companies has increased to six, indicating a growing trend of long-term capital inflow into the market[13]