Market Overview - The domestic market indices showed positive performance with the Shanghai Composite Index closing at 4,108.57, up by 0.64%, and the Shenzhen Component Index at 14,088.84, up by 1.23% [4]. Industry Commentary: Communication Sector - Nvidia reported FY2026 Q4 earnings that exceeded market expectations, with revenue reaching $68.1 billion, a year-on-year increase of 73% and a quarter-on-quarter increase of 20%. Net profit was $43 billion, up 94% year-on-year and 35% quarter-on-quarter [7]. - The data center revenue for Nvidia in Q4 was $62.3 billion, showing a year-on-year growth of 75% and a quarter-on-quarter growth of 22%. The hyperscale cloud service providers remain the largest customer segment [7]. - Nvidia's guidance for Q1 FY2027 is approximately $78 billion in revenue, surpassing previous market expectations. The growth is driven by the demand for AI programming workflows and the introduction of new products [7]. Industry Trends: Satellite Internet and AI - The upcoming MWC 2026 will focus on smart infrastructure, AI empowerment, and satellite communication, highlighting the increasing importance of satellite technology in modern warfare and communication [8]. - The DeepSeek v4 model is expected to launch soon, featuring significant improvements in multimodal capabilities and optimization for domestic chip manufacturers, indicating a strong position in the inference market [8]. Investment Recommendations - Suggested companies to watch include those involved in CPO/NPO optical engines such as Zhongji Xuchuang and NewEase, as well as domestic computing firms like Huafeng Technology and Cambricon [11]. - The overall market saw an increase, with the Shenwan Communication Index rising by 4.76% during the week, indicating a positive trend in the sector [11]. International Sportswear Brands Revenue Outlook - For the fiscal year 2026, various international sportswear brands are projected to see revenue growth, with Amer Sports expecting a 16%-18% increase, and Asics forecasting a 17.2% growth [12][13]. - Adidas anticipates high single-digit growth, while Columbia expects a modest increase of 1%-3% [13].

Investment Rating - The report maintains an "Outperform" rating for the communication industry, indicating an expected performance exceeding the benchmark index by over 10% [1]. Core Insights - Nvidia's FY2026 Q4 results exceeded market expectations, with revenue reaching $68.1 billion, a year-on-year increase of 73% and a quarter-on-quarter increase of 20%. Net profit was $43 billion, up 94% year-on-year and 35% quarter-on-quarter. The data center revenue for Q4 was $62.3 billion, with year-on-year growth of 75% and quarter-on-quarter growth of 22% [2][11]. - The growth in Nvidia's data center revenue is driven by the NVLink computing architecture and the development of Ethernet and InfiniBand platforms. Companies in the copper connection and 1.6T optical module supply chain are expected to see significant quarter-on-quarter growth in Q4 2025 and Q1 2026 [2][11]. - Nvidia's guidance for Q1 FY2027 is approximately $78 billion in revenue, surpassing previous market expectations. The report anticipates that the growth of AI programming workflows will be a major driver for token growth in the near term [2][11]. Summary by Sections Industry Trends - Nvidia is expected to unveil its latest roadmap for CPO, LPU, and Feynman at GTC2026 in mid-March, which could catalyze the overseas computing power sector. The Rubin platform is anticipated to include various SKU products, enhancing capabilities in computing, HBM, and optical connections [2][12]. - The upcoming MWC 2026 will focus on AI-enabled enterprises, 6G, optical communication, and satellite communication, highlighting the importance of satellite technology in modern warfare [3][14]. Market Overview - The overall market saw an increase during the week of February 23-27, 2026, with the Shenwan Communication Index rising by 4.76%. The top-performing sectors included IDC (+36.44%), connectors (+11.39%), and optical cables (+9.11%) [5][15]. - Notable stock performances included Runze Technology and Ruikeda, with increases of 35.15% and 30.71%, respectively. Conversely, stocks like Wangsu Technology and Changxin Bochuang experienced declines [5][28]. Recommended Companies - The report suggests focusing on companies in various sectors: - CPO/NPO optical engines: Zhongji Xuchuang, Xinyi Sheng, Tianfu Communication - In-cabinet optical passive devices: Taicheng Light, Tianfu Communication - Domestic computing power: Huafeng Technology, Huagong Technology - Satellite internet: Aerospace Electronics, Xinke Mobile [5][15].

Investment Rating - The report maintains an "Outperform" rating for the communication industry, indicating an expected performance exceeding the benchmark index by more than 10% [1][38]. Core Insights - Nvidia's FY2026 Q4 results exceeded market expectations, with revenue reaching $68.1 billion, a year-on-year increase of 73% and a quarter-on-quarter increase of 20%. Net profit was $43 billion, up 94% year-on-year and 35% quarter-on-quarter. The data center revenue for Q4 was $62.3 billion, with year-on-year growth of 75% and quarter-on-quarter growth of 22% [4][14]. - The report highlights the anticipated growth in the AI programming workflow, driven by the increasing demand for tokens, which is expected to lead to significant market expansion in the B2B and G2B sectors [14]. - Upcoming announcements at GTC2026 regarding Nvidia's CPO, LPU, and Feynman roadmap are expected to catalyze developments in the overseas computing power sector [5][15]. Summary by Sections Industry Trends - The communication industry has shown a strong market performance over the past year, with significant growth in specific segments such as IDC, connectors, and optical cables [2][8]. - The overall market saw an increase, with the Shenwan Communication Index rising by 4.76% during the week of February 23 to February 27, 2026 [18]. Key Companies to Watch - Companies recommended for attention include: - CPO/NPO Optical Engines: Zhongji Xuchuang, New Yisheng, Tianfu Communication, Huanxu Electronics, Yuanjie Technology - In-cabinet Optical Passive Devices: Taicheng Light, Tianfu Communication, Zhishang Technology, Weike Technology, Changxin Bochuang, Shijia Photon, Hengdong Light - Domestic Computing Power: Huafeng Technology, Huagong Technology, Cambrian, Moore Threads, Muxi Shares, Tianshu Zhixin, Shengke Communication - Satellite Internet: Aerospace Electronics, Xinke Mobile, Fenghuo Communication, Shanghai Huanxun, Changjiang Communication, Zhimingda, Electric Science Blue Sky, Electric Science Chip [8][18]. Market Performance - The report notes that the top three performing segments for the week were IDC (+36.44%), connectors (+11.39%), and optical cables (+9.11%) [18][20]. - Individual stock performance highlighted significant gains for companies such as Runze Technology (+35.15%), Ruikeda (+30.71%), and Gaolan Shares (+24.31%) [18][31].

Core Insights - DeepSeek has released a new paper titled "Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models," in collaboration with Peking University, introducing a new module called Engram to enhance the efficiency of large language models [1][3]. Group 1: Research Overview - The current approach to sparsity in large language models primarily relies on Mixture of Experts (MoE) for conditional computation, but existing Transformer architectures lack a native knowledge retrieval mechanism [3][8]. - DeepSeek proposes conditional memory as a complementary dimension to MoE, introducing the Engram module to facilitate efficient knowledge retrieval with O(1) time complexity [8][9]. Group 2: Engram Module Implementation - The Engram module has been implemented and made available on GitHub, allowing for community engagement and further development [4][5]. - Engram separates static memory storage from dynamic computation processes within the Transformer architecture, enhancing overall model performance [10][12]. Group 3: Performance Metrics - Engram has shown significant improvements in various benchmarks, including a +3.4% increase in MMLU accuracy and a +4.0% increase in CMMLU accuracy, as well as notable gains in general reasoning tasks [9][28]. - The architecture allows for better long-context retrieval capabilities, with accuracy in Multi-Query NIAH increasing from 84.2 to 97.0 [9]. Group 4: Experimental Results - DeepSeek trained four models: Dense-4B (4.1 billion parameters), MoE-27B (26.7 billion), Engram-27B (26.7 billion), and Engram-40B (39.5 billion), all under the same training conditions [25][27]. - The sparse architectures (MoE-27B, Engram-27B/40B) outperformed the dense model (Dense-4B) across all benchmarks, demonstrating superior scalability [28][30]. Group 5: Memory and Computation Decoupling - Engram's deterministic retrieval mechanism allows for the decoupling of parameter storage from computational resources, enabling efficient scaling without increasing computational costs [15][17]. - The architecture supports a multi-level cache hierarchy, optimizing memory access and reducing latency [18]. Group 6: U-Shaped Scaling Law - DeepSeek identified a U-shaped scaling law for optimal allocation between MoE and Engram, suggesting that a balanced distribution of sparse parameters leads to improved performance [19][24]. - The optimal allocation ratio was found to be around 20%-25% of the sparse parameter budget for Engram, confirming the structural complementarity between the two modules [23][24].

Core Insights - DeepSeek has introduced a new research paper titled "Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models," in collaboration with Peking University, focusing on enhancing large language models (LLMs) through a novel approach to memory and computation [1][2]. Group 1: Research Background and Problem Statement - Current large language models primarily utilize Mixture of Experts (MoE) for sparsity, known as "conditional computation," but lack an inherent knowledge retrieval mechanism, leading to inefficient simulation of retrieval behavior [2][8]. - DeepSeek proposes "conditional memory" as a complementary approach to MoE, introducing a new module called Engram to address this limitation [3][8]. Group 2: Engram Module and Its Implementation - The Engram module has been made available on GitHub, allowing for community engagement and further development [4]. - Engram modernizes classic n-gram embeddings to achieve knowledge retrieval in O(1) time complexity, enhancing the efficiency of memory access [8][10]. - The module separates static knowledge storage from dynamic computation processes, enhancing the overall architecture of the Transformer network [12][14]. Group 3: Performance and Efficiency - DeepSeek has expanded Engram to a scale of 27 billion parameters, demonstrating significant performance improvements over pure MoE baseline models under equivalent parameter and FLOPs conditions [10][37]. - Engram has shown notable gains in knowledge retrieval tasks, with improvements such as +3.4 in MMLU and +4.0 in CMMLU, as well as enhanced general reasoning capabilities [10][37]. - The architecture allows for efficient memory access without additional performance overhead, supporting prefetching from host memory during runtime [11][18]. Group 4: Sparsity Distribution and Optimal Allocation - DeepSeek formalized a U-shaped expansion rule to characterize the optimal trade-off between neural computation (MoE) and static memory (Engram) [9][22]. - The research indicates that a balanced allocation of approximately 20%-25% of sparse parameter budget to Engram yields optimal performance, confirming the structural complementarity between the two modules [27][29]. Group 5: Experimental Results - Four models were trained: Dense-4B, MoE-27B, Engram-27B, and Engram-40B, all under identical training conditions [34][35]. - Sparse architectures consistently outperformed the dense model across various benchmarks, with Engram-27B achieving significant improvements over MoE-27B in multiple tasks [37]. - Engram-40B further reduced pre-training loss and improved performance on most benchmarks, indicating that memory capacity has not yet reached saturation [38]. Group 6: Long Context Training - Engram's architecture has been validated for its structural advantages in long-context tasks, demonstrating significant performance gains in global context retention [40][41]. - Controlled experiments revealed that Engram outperforms MoE in complex retrieval tasks, showcasing its inherent architectural superiority [45].