In the year since DeepSeek shocked the world with a whizzy new AI model, China’s clout in the tech has only grown. Turning a profit, however, is proving difficult https://t.co/mKFC9aiAdi ...

BREAKING: Grok has officially surpassed DeepSeek in generative AI web traffic and is climbing global AI rankings fast. 🚀With monthly updates, Grok is now:#1 in several countriesTop 10 overall on App Store & Play StoreTop 1–5 in Productivity worldwide https://t.co/R2rIMssvW4 ...

Last year DeepSeek became a byword for a new wave of Chinese innovation in artificial intelligence. There are three reasons Europe should embrace it https://t.co/jc7i5kk6RlIllustration: Simon Bailly https://t.co/V4ox98dMYK ...

Group 1 - The core viewpoint of the news highlights the ongoing growth and investment opportunities in the artificial intelligence (AI) sector, driven by various industry developments and events [1][2] - The China Securities AI Theme Index (930713) shows mixed performance among its constituent stocks, with Zhongke Xingtou leading with a 14.50% increase, while AI ETF (512930) is currently priced at 2.37 yuan [1] - The third National AI Application Scenario Innovation Challenge is set to take place from January 23 to 25 in Suzhou, indicating a focus on innovation and application in the AI field [1] - Zhejiang Wenlian reported that its AI programmatic advertising tool "Paizhi" has achieved a cumulative consumption of over 250 million yuan by 2025, reflecting a year-on-year growth of 500% [1] - DeepSeek plans to release its V4 model in February 2026, which is expected to surpass the programming capabilities of Claude and GPT series, potentially reshaping the global AI market competition [1] Group 2 - The AI Theme Index includes 50 listed companies that provide essential resources, technology, and application support for AI, reflecting the overall performance of AI-related securities [2] - As of December 31, 2025, the top ten weighted stocks in the AI Theme Index account for 58.08% of the index, with companies like Zhongji Xuchuang and Xinyi Sheng among the leaders [2] - The AI ETF closely tracks the performance of the AI Theme Index, with various connection options available for investors [2]

Core Insights - The article discusses the evolving landscape of AI at the Davos conference, highlighting a shift from uncertainty to a more determined outlook amidst geopolitical tensions. The focus has moved from the capabilities of open-source models to deeper strategic considerations in AI development [3][4]. Group 1: AI Discussion Focus - The first key focus is Google's strong comeback and its comprehensive AI ecosystem, which has significantly reduced its inference costs to less than 30% of OpenAI's, providing a competitive edge through vertical integration [5][6]. - The second focus is the rise of open-source ecosystems and smaller models, with companies leveraging high-quality data for fine-tuning, leading to lower costs and enhanced data privacy for B2B applications [6][7]. - The third focus is the growing concern among European companies regarding data sovereignty, prompting a demand for localized AI solutions due to geopolitical tensions with the U.S. [7][8]. Group 2: AI Implementation Challenges - The competition in AI has shifted from model capabilities to cost, energy consumption, and industry integration, with smaller models gaining traction for their practicality and economic advantages in localized deployments [8][9]. - Energy supply has emerged as a critical bottleneck, with concerns about the aging U.S. power grid and the need for tech companies to invest in energy infrastructure to support AI operations [9][10]. - The integration of AI into industries such as healthcare and finance is accelerating, with significant investments aimed at developing AI-driven treatment solutions for diseases like Parkinson's and Alzheimer's [10][11]. Group 3: China's Competitive Advantages - China's strengths in infrastructure, particularly in electricity and robotics, are becoming significant competitive advantages in the global AI landscape, especially as U.S. companies face energy challenges [12][13]. - The cost efficiency of Chinese robotics companies, exemplified by firms like Yushutech, positions them favorably in the market, with costs potentially being one-tenth of their U.S. counterparts [13][14]. - China's innovation in pharmaceuticals is gaining recognition, with multinational companies increasingly acquiring Chinese biotech firms, indicating a shift in the global competitive landscape [14][15].

Core Insights - The storage sector is experiencing a "perfect storm," with SanDisk's stock price increasing over 100%, driven by a value reassessment triggered by advancements in AI architecture [1][11] - Storage is transitioning from a cost item to a core production element for AI, as evidenced by developments from NVIDIA and DeepSeek [1][10] Group 1: Technological Developments - NVIDIA's CEO Jensen Huang introduced the concept of Inference Context Memory Storage (ICMS) at CES 2026, highlighting that context is becoming a new bottleneck for AI rather than computing power [2][3] - The new DGX Vera Rubin NVL72 SuperPOD architecture includes dedicated storage racks for inference context, significantly increasing NAND requirements [2][3] - DeepSeek's Engram model allows NAND to be used as slow memory, enabling deterministic memory access and reducing latency issues compared to HBM [4][5][8] Group 2: Market Implications - The global NAND market, with an annual demand of approximately 1.1–1.2 ZB, is expected to see nearly 10% structural growth driven by AI infrastructure rather than traditional consumer electronics [3][11] - NAND's role is evolving from merely cold data storage to being integrated into a tiered memory system, acting as "slow RAM" for AI applications [8][9] - The combination of BlueField DPU and NAND offers a cost-effective solution for long-term memory needs in AI agents, decoupling storage demand from traditional computing resources [9][10] Group 3: Strategic Value of NAND - The strategic value of NAND is being re-evaluated as it becomes indispensable in AI architectures, leading to a potential shift in pricing logic [11] - Analysts suggest that the developments in NAND technology represent a path to achieving more efficient storage-computing collaboration, which may be more cost-effective than merely expanding computing power [8][9][11]

最令她意外的，是地缘政治下欧洲企业的"数据主权"焦虑。 "昨晚回到酒店11点半，还是比较早的。"这是Fusion Fund 创始合伙人张璐在达沃斯的第三天，每天睡眠不足5小时，行程表被无数场圆桌讨论、闭门早餐 会、深夜晚宴和远程会议填满。 "你可以明显感到行业的紧迫感。"张璐在接受第一财经记者独家采访时说。与去年的迷茫不同，尽管特朗普带来的地缘政治不确定性依然像乌云笼罩，但今 年的达沃斯多了一份笃定。大家不再纠结于"世界会怎样"，因为答案已经摆在眼前——世界已经不可逆地进入了一个"多体系并行"的常态。 张璐回忆，去年参加达沃斯，DeepSeek曾给全球科技界带来了一场"反直觉"的震撼——这家中国创业公司用极低的成本和开源的姿态，打破了原本被认为坚 不可摧的算力与模型垄断，成为当时那场关于"AI霸权"讨论中最具话题性的变量。 今年她捕捉到，大家对于AI讨论的重心已经从去年对开源模型的惊叹，转向了更冷酷、更务实的深层博弈。从技术路径、成本竞争，到能源瓶颈，再到人 类未来向何处去的思考，AI讨论进入"深水区"。 Kway He the state 75 15 14 14 ning and and arat - St ...

Summary of Conference Call on DeepSeek's New Model Industry Overview - The conference focused on the advancements in the artificial intelligence (AI) sector in China, particularly through the innovative n-gram hybrid architecture developed by DeepSeek, which enhances large model inference efficiency and reduces infrastructure costs [1][2][3]. Key Points and Arguments Innovations in AI Architecture - DeepSeek's n-gram module separates storage and computation, alleviating AI computational bottlenecks and reducing reliance on high-bandwidth memory (HBM), leading to significant improvements in inference capability and cost efficiency [2][3][4]. - The next-generation models will continue to leverage the n-gram architecture to enhance encoding and logical reasoning capabilities [2][3]. Impact on the AI and Semiconductor Industry - The efficiency-driven innovations in AI are narrowing the gap between Chinese AI companies and international leaders, creating positive investment opportunities in the AI and semiconductor sectors [2][4]. - The n-gram architecture's design allows for improved model efficiency through conditional memory and scalable lookup, which is crucial for China's strategy to catch up with the U.S. in AI technology [2][4][10]. Financial Analysis and Valuation - Discussions included the valuation models for companies like NATechnology Group, JCET Group, and AMEC, with key assumptions such as WACC of 10.1%, mid-term growth rate of 14%, and perpetual growth rate of 5% for NATechnology [12][13][14]. - Risks to these valuations include potential downturns in semiconductor investments, market share losses, and weak downstream demand leading to chip oversupply [12][13][14]. Additional Important Insights - The n-gram technology allows for efficient use of DRAM, requiring only about 200GB to achieve good inference performance, significantly lowering hardware barriers and costs compared to traditional systems [17]. - The architecture's ability to optimize GPU utilization and system innovation enables Chinese AI companies to reduce dependency on the latest hardware while still competing with top models globally [10][21]. - Comparisons between Chinese models like DeepSeek v3.2 and ChatGPT 5.2 show that while Chinese models excel in multi-modal reasoning and long-context processing, they still lag in task coverage compared to ChatGPT [21]. Conclusion - The advancements in n-gram technology and its implications for the AI and semiconductor industries highlight a transformative period for Chinese AI companies, emphasizing efficiency and innovation over sheer computational power. The financial outlook for related companies remains cautiously optimistic, contingent on market dynamics and technological progress [2][12][13][14].

Investment Rating - The report maintains an "Outperform" rating for the computer industry, indicating a positive outlook compared to the market [1]. Core Insights - The acceleration of AI applications is expected to significantly reshape the global AI competitive landscape, particularly with the anticipated release of DeepSeek's V4 model [7]. - The Chinese AI cloud market is projected to reach 51.8 billion yuan by 2025 and 193 billion yuan by 2030, with Alibaba aiming to capture 80% of the market's incremental growth in 2026 [7]. - The report highlights the ongoing trend of domestic chip manufacturers gaining market share due to restrictions on foreign competitors, particularly Nvidia's H200 [7]. Summary by Sections Industry Data - From January to November 2025, the software industry revenue reached 13.98 trillion yuan, growing by 13.3% year-on-year, marking a continuous recovery over nine months [13]. - The IC design sector showed the highest growth rate at 16.5%, outperforming the overall software industry growth [18]. AI Developments - Major AI models such as OpenAI's GPT-5 and DeepSeek's V3.2 are leading the market, with DeepSeek's models expected to challenge established players significantly [41][46]. - The report notes that the trend of using domestic chips for training large models is expected to gain momentum in 2026, with DeepSeek already optimizing its models for compatibility with domestic chips [66]. Domestic Market Trends - The report emphasizes the increasing number of devices running on Huawei's HarmonyOS, which has surpassed 36 million, indicating a strong push towards domestic technology adoption [7]. - The AI cloud market is becoming increasingly competitive, with Alibaba and Volcano Engine emerging as the two dominant players [7]. Investment Opportunities - The report suggests focusing on companies like Runze Technology, Sugon, and Zhongke Shuguang, which have significant roles in the AI infrastructure and domestic chip development [7]. - It also highlights the potential of companies like Changxin Technology and Chipone Semiconductor, which are actively pursuing IPOs [7].

Core Insights - DeepSeek is revolutionizing AI scalability by utilizing a hybrid architecture that replaces scarce high-bandwidth memory (HBM) with more cost-effective DRAM through an innovative module called "Engram" [1][3][5] Group 1: Engram Module and Conditional Memory - The Engram module introduces "Conditional Memory," separating static knowledge storage from dynamic reasoning, which significantly reduces reliance on expensive HBM [3][5] - This architecture allows for efficient retrieval of basic information without overloading HBM, thus freeing up capacity for more complex reasoning tasks [3][5] Group 2: Economic Impact on Infrastructure - The Engram architecture reshapes hardware cost structures by minimizing HBM dependency, potentially shifting infrastructure costs from GPUs to more affordable DRAM [5][6] - A 100 billion parameter Engram model requires approximately 200GB of system DRAM, indicating a 13% increase in the use of commodity DRAM per system [5][6] Group 3: Innovation Driven by Constraints - Despite limitations in advanced computing power and hardware access, Chinese AI models have rapidly closed the performance gap with global leaders, demonstrating "constraint-induced innovation" [6][7] - DeepSeek's advancements suggest that future AI capabilities may rely more on algorithmic and system-level innovations rather than merely increasing hardware resources [6][7] Group 4: Future Outlook - The upcoming DeepSeek V4 model is expected to achieve significant advancements in encoding and reasoning, potentially running on consumer-grade hardware like the RTX 5090 [7] - This development could lower the marginal costs of high-level AI inference, enabling broader deployment of AI applications without the need for expensive data center-grade GPU clusters [7]