Investment Rating - The report maintains an "Overweight" rating for the technology and computer sectors [6]. Core Insights - The demand for computing power is expected to grow non-linearly due to the rise of Agentic AI, with token usage projected to increase by over 10 times, leading to a corresponding increase in computing power demand by over 100 times [1][90]. - The report highlights three scaling laws: pre-training scaling, post-training scaling, and inference scaling, which collectively indicate that the demand for computing power will continue to grow significantly [10][11]. - The relationship between token consumption and computing power demand is not linear, with a 10-fold increase in token usage potentially resulting in a 100-fold increase in required computing power [60][90]. Summary by Sections Token Demand and Computing Power - Token usage and computing power demand are expected to grow non-linearly, with the complexity of inference processes requiring significantly more computing resources as token usage increases [1][60]. - The report cites Huang Renxun's statement that a 10-fold increase in token volume could lead to a 100-fold increase in computing power requirements due to the complexity of inference processes [1][60]. Scaling Laws - The report discusses three scaling laws: pre-training scaling, post-training scaling, and inference scaling, emphasizing that the market may be underestimating the future demand for computing power due to concerns about the peak of pre-training scaling [10][11]. - Inference scaling is particularly important for improving model performance on difficult problems, which is essential for the development of Agentic AI [15][19]. Agentic AI and Token Consumption - The report identifies Deep Research as a significant driver of token consumption, with estimates suggesting that its token usage could be up to 50 times that of a single chat interaction [3][50]. - The complexity of tasks handled by Agentic AI leads to higher token consumption, with the potential for token usage to exceed 100 times that of traditional chat interactions in more complex scenarios [57][58]. Future Outlook - The report concludes that the future demand for computing power will be driven by the dual factors of increasing token usage and the complexity of inference tasks, indicating a broad space for growth in computing power demand [89][90].

Core Viewpoint - The release of Grok 4 by XAI demonstrates significant advancements in reasoning capabilities for professional disciplines and complex tasks, indicating potential applications in high-value scenarios and driving demand for AI infrastructure and computing power [1][2]. Group 1: Product Release and Pricing - Grok 4 has been officially launched, featuring two versions: Grok 4 and Grok 4 Heavy, with enhanced performance in professional tasks [2]. - The pricing for the B-end API is set at $3 per million tokens for input and $15 per million tokens for output, approximately 50% higher than the previous version [2]. - C-end users can access Grok 4 for a subscription fee of $30 per month, while the high-performance Grok 4 Heavy version costs $300 per month [2]. Group 2: Performance Enhancements - Grok 4 significantly outperforms previous state-of-the-art models in reasoning tasks, achieving a 26.9% accuracy rate without tools and 41.0% with tools on the Humanity's Last Exam (HLE) test set, with potential to reach 50.7% through increased reinforcement learning (RL) computation [3]. - In the Vending-Bench test, Grok 4 scored twice as high as the second-place model, Claude Opus 4, indicating its capability in solving complex real-world problems [3]. - Grok 4 Heavy excelled in several academic knowledge tests, achieving near-perfect scores in AIME25 and HMMT25 [3]. Group 3: Computational Demand and Technical Innovations - The training volume for Grok 4 has increased by 100 times compared to Grok 2, and the computational load for post-training reinforcement learning has increased tenfold compared to Grok 3 [4]. - Grok 4 Heavy has validated the effectiveness of increased RL computation in enhancing model performance, demonstrating superior cost-effectiveness in reasoning compared to all previous models [4]. - Key engineering innovations include the importance of tool usage in improving reasoning performance and the development of reliable reward signal schemes in post-training reinforcement learning [4]. Group 4: Future Developments and Multimodal Capabilities - The new voice assistant, Eve, has reduced conversation latency by 50% and increased daily user engagement by tenfold, showcasing advanced conversational abilities [5]. - There are plans to enhance visual understanding and generation capabilities in upcoming updates, with a focus on multimodal intelligence [5]. - Future releases include a code model in August, a multimodal agent in September, and a video generation model in October [5].

Summary of Conference Call on AI Industry Development Industry Overview - The conference call primarily discusses the advancements in the AI industry, focusing on the performance and features of the GROX4 model and the anticipated release of GPT-5. [1][2][4] Key Points and Arguments GROX4 Model Advancements 1. **Significant Improvement in Reasoning Ability**: GROX4 achieved a score of 50 in the Humans Last Examination (HLE), surpassing OpenAI's score of 23, and excelled in the US Olympic Math Competition with scores of 97 and 90 in HNMT and USAMO respectively, indicating a doubling of previous performance levels. [3][4] 2. **Parameter Optimization and Efficiency**: The model reduced its parameter count by 40% through sparse activation strategies, using only 1.7 trillion tokens compared to GROX3's 2.7 trillion tokens while significantly enhancing performance. [3][4] 3. **Multimodal Fusion and Real-time Search**: GROX4 integrates audio, images, real-time search, and tool invocation, allowing it to handle complex tasks more intelligently and support real-time internet functionality. [3][4] 4. **High API Pricing**: The API pricing for GROX4 is set at $3 per million tokens for input and $15 per million tokens for output, reflecting a significant increase in costs due to performance enhancements. [1][6] GPT-5 Expectations 1. **Release Timeline**: GPT-5 is expected to be released between late July and September 2025, with a focus on deep multimodal integration, including text-to-image, text-to-video, and audio interaction capabilities. [5][26] 2. **Technical Improvements**: The model aims to enhance agent functionalities and address shortcomings in product experience, although it may face challenges in achieving satisfactory benchmark results. [5][26] Market Trends and Implications 1. **Growing Demand for High-Performance Computing**: The rapid development of AI large models and reinforcement learning technologies is driving an increasing demand for computational resources, as evidenced by Nvidia's market valuation surpassing significant thresholds. [2][8][19] 2. **Impact on AI Industry Structure**: The introduction of Grok's innovative training methods may alter the division of labor within the AI industry, potentially squeezing out smaller startups while creating new opportunities for those with unique data or capabilities. [11][12] 3. **Future GPU Demand**: The AI industry's growth is expected to lead to exponential increases in GPU demand, with projections indicating a need for up to 1 million high-performance GPUs in the coming years. [19][20] Additional Insights 1. **Challenges in Programming Capabilities**: Despite high benchmark scores, GROX4's programming capabilities may not meet expectations due to potential contamination in training data and limitations in user interaction history. [14][15] 2. **Pricing Strategy Justification**: The high subscription fee of $300 per month for GROX4 reflects both confidence in its capabilities and cost considerations, although it may not significantly outperform other leading models for average users. [15][16] 3. **Potential for New Opportunities**: The evolving technical paradigms in AI may create new opportunities, particularly in fields like scientific research, where AI could lead to breakthroughs in areas such as drug development and DNA research. [13][12] Conclusion The conference call highlights significant advancements in AI technology, particularly with the GROX4 model, while also addressing the anticipated developments with GPT-5. The ongoing demand for computational resources and the potential restructuring of the AI industry present both challenges and opportunities for various stakeholders.

Group 1 - The diesel generator sector is undergoing adjustments primarily due to the impact of chip bans, leading to lower-than-expected capex intensity and pace from domestic manufacturers [1] - Nvidia is developing a new chip, B30, specifically for the Chinese market, with plans to produce over 1 million units this year [1] - Despite slow progress in data center construction in Q2, overall demand remains strong, extending the industry's prosperity until at least the end of next year [2] Group 2 - Major manufacturers plan to initiate bidding for diesel generators for 2026 in advance, with both the bidding volume and prices expected to exceed previous expectations from April to May [3] - Weichai Power (02338) anticipates selling over 8,100 units of its M series large-bore engines in 2024, with overseas sales accounting for 62% and a 20% year-on-year revenue increase, driven by a 148% surge in data center product sales [4] - Chongqing Machinery and Electric (02722) projects a net profit of approximately 431 million yuan for 2024, reflecting a 42% increase, benefiting from the sustained high demand for high-end diesel generators due to ongoing AIDC and data center construction [4]

Group 1 - The core viewpoint of the article highlights the strong growth in AI computing infrastructure globally, with significant investments from companies like Nvidia and Oracle, indicating a robust demand for cloud services and AI capabilities [1] - Nvidia plans to establish 20 new "AI factories" in Europe, aiming for a tenfold increase in AI computing capacity within two years, equipped with 10,000 GPUs, while also enhancing collaborations in quantum computing and industrial AI [1] - Oracle's cloud infrastructure revenue has increased by 52% year-on-year, with expectations of over 70% growth in the fiscal year 2026, alongside a capital expenditure projection of $25 billion, reflecting strong demand in the cloud services sector [1] Group 2 - The communication industry, particularly in the optical module and optical device sectors, has shown a notable performance with a weekly increase of 2.07%, driven by the demand for AI computing [1] - The upcoming 2025 Mobile World Congress (MWC) will focus on themes such as 5G integration and AI, further promoting innovation within the industry [1] - The communication ETF (515880) tracks the communication equipment index (931160), which includes listed companies involved in communication network equipment, terminal devices, and related services, reflecting the overall performance of the communication equipment sector [1]

Group 1 - The software industry in China showed a positive performance in the first four months of 2025, with software business revenue reaching 42,582 billion yuan, a year-on-year increase of 10.8%, and total profit of 5,075 billion yuan, up 14.2% year-on-year [1] - The Ministry of Industry and Information Technology (MIIT) plans to accelerate the development of open-source infrastructure, particularly in foundational and industrial software, as 97% of global developers and 99% of enterprises are using open-source software [1] - A large-scale equipment update project in the manufacturing sector is expected to stimulate a market exceeding 50 trillion yuan, promoting technological upgrades in traditional industries [1] Group 2 - Industrial software is becoming a core tool for optimizing manufacturing management processes and changing production models as China transitions to smart manufacturing [2] - By 2027, MIIT aims to update approximately 2 million sets of industrial software and 800,000 operating systems across key industries such as petroleum, chemicals, and aviation [2] - The industrial software market in China reached 282.4 billion yuan in 2023, with a five-year compound annual growth rate of 13.8%, and is projected to grow to 333.2 billion yuan in 2024, reflecting an 18.0% year-on-year increase [2] - The National Development and Reform Commission has included industrial software upgrades in the "Two New" policy support scope, indicating expanded policy support to encourage enterprise investment [2] - Upgrading industrial software can directly enhance enterprise efficiency, reduce operational costs, and strengthen competitiveness, although there are still shortcomings in supporting the digital transformation of manufacturing [2] - The software ETF (515230) tracks the software index (H30202), which reflects the overall performance of listed companies in the software industry, characterized by significant growth and technological innovation [2]

近日，北京昆仑联通科技发展股份有限公司（以下简称"昆仑联通"）对北交所第二轮审核问询函作出回复，并披露了财务数据更新版的第一轮问询回复。其 中该公司与蔚来汽车（09866.HK）及字节跳动2个核心客户之间合作的变化，引发市场关注。 | 代码 | 874435 | | --- | --- | | 公司全称 | 北京昆仑联通科技发展股份有限公司 | | 审核状态 | 已问询 | | 保存机构 | 东方证券股份有限公司 | | 会计师事务所 | 中汇会计师事务所(特殊普通合伙) | | 律师事务所 | 北京市竞天公诚律师事务所 | | 评估机构 | | 来源：北交所 昆仑联通成立于1998年，是一家专业的IT基础架构解决方案提供商，主要为客户提供H3C服务器、MS-O365、HP服务器等产品。问询函显示，该公司与蔚 来汽车首次合作的时间为2015年，主要向后者提供数据中心以及数字化办公相关业务。二者的"牵手"也被该公司看作是其拓展新能源汽车领域客户的成功案 例。 招股书显示，2021年至2023年，随着新能源汽车行业爆发，蔚来汽车营收规模持续增长。受益于此，昆仑联通对蔚来汽车的销售收入从5212.97万元增长至 1 ...

Group 1: Key Insights on Xinjiang Coal Industry - Xinjiang has abundant coal reserves with a resource volume of 2.19 trillion tons, accounting for 39% of the national total, and confirmed coal resources of 450 billion tons [3][12] - The coal production in Xinjiang is projected to reach 900-1,000 million tons by 2030, with an expected export of 250 million tons [3][12] - The railway infrastructure in Xinjiang is undergoing significant development, with the completion of the Lin-Ha backbone channel by 2027 expected to add an export capacity of 200 million tons [3][12] Group 2: Investment Opportunities in Xinjiang Coal - The coal consumption in Xinjiang is anticipated to grow significantly, driven by the construction of large-scale thermal power plants and coal chemical projects [3][12] - Key companies to watch for potential growth in production capacity in Xinjiang include Tebian Electric Apparatus, Hubei Yihua, and Yanzhou Coal Mining [3][12] Group 3: Insights on Logistics and Robotics Industry - The logistics automation and robotics sector is expanding, with applications across various scenarios including factory logistics, commercial delivery, and consumer logistics [19] - The core drivers for the adoption of robotics in logistics are improving operational efficiency and reducing labor costs, particularly in the express delivery sector [19] - Companies like Dematic Technology and others are actively engaging in strategic partnerships to enhance logistics automation capabilities [19]

Group 1 - The core viewpoint of the news highlights that the company Shijia Photon is adapting to international trade policy changes with limited impact through overseas capacity layout and diversification strategies [1] - In the first quarter, the company's gross margin significantly increased due to product structure upgrades, process optimization, yield improvements, and customer structure optimization [1] - The company has made progress in data centers, access networks, and laser radar and sensing markets, aiming to deepen process advantages and accelerate product industrialization [1] Group 2 - MPO product capacity layout is globalized, while MT connectors ensure supply stability through investments in suppliers and long-term agreements [1] - The existing production lines for WG products meet order demands, with future capacity adjustments based on market conditions [1] - Revenue and profit are expected to continue growing in the first quarter of 2025, driven by rapid growth in computing power demand and the data communication market [1] Group 3 - The construction of the Thailand production base is progressing steadily, with core team formation and skills training meeting expectations, leading to continuous capacity enhancement [1] - The company maintains a cautiously optimistic attitude towards the market and operational conditions in 2025 [1]

Group 1: Shijia Photon - The company reported that changes in international trade policies have limited impact, as it is addressing tariff policies through overseas capacity layout and diversification strategies [1] - In Q1, the gross margin significantly increased due to product structure upgrades, process optimization, yield improvements, and customer structure optimization [1] - The company is making progress in data centers, access networks, and laser radar and sensing markets, with plans to deepen process advantages and accelerate product industrialization [1] - Global capacity layout for MPO products and stable supply ensured for MT connectors through investments and long-term agreements [1] - The existing production line for WG products meets order demands, with future capacity adjustments based on market conditions [1] - Revenue and profit are expected to continue growing in Q1 2025, driven by rapid growth in computing power demand and the data communication market [1] - The construction of the Thailand production base is progressing steadily, with core team formation and skills training meeting expectations [1] - The company maintains a cautiously optimistic outlook for the market and operational conditions in 2025 [1] Group 2: Canrui Technology - The company expects a 24.36% year-on-year increase in revenue for 2024, but net profit is projected to turn negative due to intense market competition, high R&D expenses, impairment losses, and losses in the testing business [2] - Revenue from smart sensor chips is expected to grow by 32.61%, while power management chip growth is only 5.04% [2] - The total cost of equity incentives for 2024 is projected to be 20.68 million yuan, amortized over three years [2] - In Q1 2025, revenue is expected to reach 125 million yuan, an 8.86% year-on-year increase, with gross margin rising to 28.12% [2] - The company is enhancing profit margins through internal growth and external expansion in the competitive analog chip industry [2] - A new share buyback program has accumulated the purchase of 1,155,191 shares, costing 31,898,161.44 yuan [2] Group 3: Huaru Technology - The company highlighted the characteristics of leading foreign military intelligence enterprises and introduced the XSimVerse military model with five application areas: intelligent decision-making, virtual training, digital testing, training equipment, and intelligent equipment [3] - The company emphasized the upgrade advantages from military simulation to military intelligence, forming a closed-loop system through simulation systems and corpus resources [3] - Despite facing procurement restrictions from military networks, the company can still secure orders through alternative channels and increase potential customer demand through product iterations [3] - As of December 31, 2024, the company has signed new contracts worth 376 million yuan [3] Group 4: Changsheng Fund - As of now, Changsheng Fund has an asset management scale of 86.128 billion yuan, ranking 64th out of 210 [3] - The asset management scale for non-monetary public funds is 55.698 billion yuan, ranking 67th out of 210 [3] - The fund manages 136 public funds, ranking 54th out of 210, with 22 public fund managers, ranking 63rd out of 210 [3] - The best-performing public fund product in the past year is Changsheng Urbanization Theme Mixed A, with a latest net value of 1.62 and a growth of 53.12% over the past year [3]