公司点评报告 · 计算机行业 业绩高速增长,国产化浪潮下迎发展黄金期 2025 年 2 月 28 日 海光信息(688041.SH) 推荐 维持评级 分析师 吴砚靖 ☎: 010-66568589 | 市场数据 | 2025-2-28 | | --- | --- | | 股票代码 | 688041 | | A 股收盘价(元) | 160.00 | | 上证指数 | 3.320.90 | | 总股本(万股) | 232,433.81 | | 实际流通 A 股(万股) | 88,655.72 | | 流通 A 股市值(亿元) | 1,418.49 | 相对沪深 300 表现图 2025-2-28 料来源: Wind,中国银河证券研究 网: wuyanjing@chinastock.com.cn 分析师登记编码：S0130519070001 高峰 ☎：010-80927671 网: gaofeng_yj@chinastock.com.cn 分析师登记编码：S0130522040001 研究助理:胡天昊 ☎：010-80927637 മ: hutianhao_yj @chinastock.com.cn www.chi ...

Summary of Key Points from the Conference Call Company Overview - Company: 科大讯飞 (iFLYTEK) - Industry: Artificial Intelligence (AI) Core Insights and Arguments - Collaboration with Huawei to build the first domestic "ten thousand card cluster" computing base using Huawei's Ascend chips, enhancing competitive advantage in the localization process and continuously developing self-researched large models from Spark 1.0 to 4.0 and differential reasoning models [2][3] - Focus on vertical industries such as education and healthcare, covering over 800 central state-owned enterprises and industry leaders, leading in the number and amount of bids, establishing a significant market foothold [2][7] - Projected revenue growth exceeding 15% from 2024 to 2026, with expected revenue reaching 30 billion yuan by 2026, and continuous improvement in cash flow; long-term goals include achieving 100 billion yuan in revenue and 10 billion yuan in profit, with AI enhancements expected to increase software gross margin to 10% [2][9] - The Spark X1 model demonstrates excellent mathematical reasoning capabilities, particularly in the Chinese context, performing comparably to DPC R1 in K12 and university assessments, and outperforming in primary and high school competition math, reaching an international first-tier level [2][11][12] - Accumulated a large amount of high-quality de-identified diagnostic data in the AI healthcare sector, significantly reducing medical hallucinations, and leading in scores for general auxiliary diagnosis and health consultation, with strong logical correctness, professionalism, and interpretability [2][13] Additional Important Content - iFLYTEK has transformed from a voice technology company to a comprehensive player in large models, positioning itself as a fully autonomous national team covering foundational computing, intermediate large models, and upper-layer application scenarios [3] - Despite lower stock price growth compared to the average in the computer industry, the company is expected to maintain a 15% compound growth rate over the next two years, indicating significant market value growth potential [4][18] - The company has deepened its presence in various vertical industries, including education, healthcare, telecommunications, finance, law, and government, with education being the largest revenue segment followed by healthcare [7] - The recent upgrade of the differential reasoning model showcases the company's ongoing efforts in self-research and reinforces its leading position in the commercialization of AI [10] - iFLYTEK's service robots, launched in 2023, utilize self-developed chips and algorithms, indicating a potential growth area in the service robot market [17]

Core Viewpoint - The article emphasizes maintaining long-term confidence and increasing positions in core leading companies during market adjustments, particularly in the AI and domestic sectors [2][4]. Summary by Sections Weekly Viewpoint - The market has experienced adjustments due to previous rapid increases and the disclosure of performance reports. Looking ahead, the completion of performance reports and the upcoming important policy window are expected to improve performance in Q1 2025 [2][4]. Deep Dive on DeepSeek - DeepSeek, a leader in AI equity, has made significant progress by open-sourcing various solutions, including DeepEP and DeepGEMM, which optimize both hardware and applications. The company has also implemented a "staggered pricing" strategy, significantly reducing API call prices during off-peak hours, which is anticipated to accelerate AI application development [2][3]. Global AI Industry Trends - The global AI industry is experiencing accelerated resonance, with notable advancements from domestic models like Doubao and DeepSeek. OpenAI's release of ChatGPT 4.5 has also shown significant improvements in human-like interactions, contributing to the ongoing trend in the AI sector [2]. Investment Strategy - The recommendation is to continue over-allocating resources to the AI sector, focusing on core leading companies as the industry trends evolve and mature [2].

Group 1: DeepSeek and AI Industry Transformation - DeepSeek's technological innovation, particularly the application of Scaling Law theory, significantly enhances AI model performance [1][3] - The Scaling Law theory indicates that AI model performance is proportional to the amount of parameters, data, and computation, with simultaneous improvements leading to substantial performance gains [3] - DeepSeek optimizes model performance and reduces costs, promoting AI technology applications in traditional industries such as SMEs, healthcare, and finance, thereby stimulating the growth of computing power demand [3] - DeepSeek collaborates with domestic chip manufacturers like Huawei Ascend and Haiguang to enhance the adaptability and development of domestic chips, further strengthening the autonomy of domestic computing power [3] - Through distillation technology and algorithm optimization, DeepSeek significantly reduces model storage requirements and computational load, enabling efficient inference of AI models on smart terminals like smartphones and headphones [3] - Multiple domestic smartphone manufacturers have integrated DeepSeek's AI models, with smart wearable devices (e.g., AI headphones) becoming important scenarios for edge AI applications [3] - DeepSeek-R1 represents a breakthrough in China's open-source AI field, boasting high performance, low cost, and open-source advantages, laying a foundation for future market growth with its global influence and rapidly growing user base [3] Group 2: Investment Research and Large Language Models - Automation programming plugins support multi-mode programming, file operations, command line integration, and multi-model API calls within VSCode, enabling automatic file reading, dependency installation, code execution, and error correction [4] - Large language models can transform subjective factors in investment decisions into quantifiable variables, assisting investors in conducting more efficient quantitative analysis when developing investment models [4] - The model can automatically extract market trends, industry chain information, and corporate financial data from analyst reports, providing valuable input data for quantitative investment models [5] - The model employs sentiment analysis techniques to help research personnel extract relevant emotions and viewpoints from news, social media, and reports, further optimizing the understanding of market dynamics in quantitative investment models [5] - Intelligent agents like ChatGPTTask and Operator can automate tasks such as regularly obtaining information and browsing the web, allowing research personnel to focus more on value-creating work [5] - By constructing knowledge bases, research personnel can easily extract information from historical data and reports, even obtaining relevant answers through direct inquiries [5] - For research institutions that prefer not to invest heavily in hardware and operational costs, large model API services from platforms like OpenRouter, Huoshan Engine, and Alibaba Cloud are available [5] - Tools like Ollama simplify the installation and operation processes of large models while ensuring data privacy and security [5] Group 3: Industry Trends and Future Outlook - The government emphasizes strengthening independent innovation, reflecting a high level of importance placed on technological innovation by representatives of various technology companies [5][8] - The State-owned Assets Supervision and Administration Commission (SASAC) released implementation points for the "AI+" special action, indicating continued policy support for future technological innovation, which enhances market confidence [8] - Due to the ongoing release of policy dividends and continuous deepening of industrial innovation, AI and leading domestic companies are expected to remain the main focus for future allocations [8] - Alibaba plans to invest more in cloud and AI infrastructure over the next three years than in the past decade, demonstrating its commitment and strategic layout in the AI field [8] - The demand for inference computing power is expected to grow rapidly in the short term as AI applications expand, particularly driven by the inference needs of large models, which will become a significant driving force in the computing power industry [8]

本次发行股票拟在科创板上市，科创板公司具有研发投入大、经营风险高、业 绩不稳定、退市风险高等特点，投资者面临较大的市场风险。投资者应充分了 解科创板的投资风险及本公司所披露的风险因素，审慎作出投资决定。 首次公开发行股票并在科创板上市 招股说明书 保荐人（主承销商） （中国(上海)自由贸易试验区浦明路 8 号） 广东天承科技股份有限公司 招股说明书 声明 广东天承科技股份有限公司 Guangdong Skychem Technology Co., Ltd. （珠海市金湾区南水镇化联三路 280 号） 根据《证券法》的规定，股票依法发行后，发行人经营与收益的变化，由 发行人自行负责；投资者自主判断发行人的投资价值，自主作出投资决策，自 行承担股票依法发行后因发行人经营与收益变化或者股票价格变动引致的投资 风险。 1-1-1 广东天承科技股份有限公司 招股说明书 1-1-2 发行股票类型： 人民币普通股（A 股） 发行股数： 发行 14,534,232 股，占发行后总股本比例为 25% 每股面值： 人民币 1.00 元 每股发行价格： 55.00 元 发行日期： 2023 年 6 月 28 日 拟上市的交易所和板 ...