Financial Performance - Total revenue for 2023 decreased by 10.6% to RMB 3,405.8 million from RMB 3,808.5 million in 2022[9] - Gross profit fell to RMB 1,500.8 million in 2023, down from RMB 2,542.3 million in 2022, resulting in a gross margin of 44.1%[11] - The adjusted EBITDA loss increased to RMB 4,369.0 million in 2023 from RMB 4,213.4 million in 2022[9] - Net loss for the year rose to RMB 6,494.7 million from RMB 6,093.0 million in 2022[9] - Revenue decreased by 10.6% from RMB 3,808.5 million in 2022 to RMB 3,405.8 million in 2023, primarily due to a decline in traditional AI revenue, partially offset by increases in generative AI and smart vehicle revenue[38] - Traditional AI revenue fell by 41.1% from RMB 3,121.1 million in 2022 to RMB 1,838.4 million in 2023, reflecting a significant reduction in reliance on smart city business[39] - Cost of sales rose by 50.5% from RMB 1,266.2 million in 2022 to RMB 1,905.1 million in 2023, mainly due to increased hardware costs and subcontracting service fees[40] - Gross profit decreased by 41.0% from RMB 2,542.3 million in 2022 to RMB 1,500.8 million in 2023, with gross margin dropping from 66.8% to 44.1%[41] - Annual loss increased to RMB 6,494.7 million in 2023 from RMB 6,093.0 million in 2022, as operational cost savings were insufficient to offset the decline in gross profit[52] Revenue Breakdown - Revenue from generative AI services surged by 199.9% to RMB 1,183.7 million, accounting for 34.8% of total revenue[11] - The "Absolute Shadow" smart vehicle business generated RMB 383.7 million in revenue, a 31.1% increase year-over-year[11] - Revenue from traditional AI businesses, including smart cities, declined by 41.1% to RMB 1,838.4 million, representing 53.9% of total revenue[11] - Smart vehicle revenue increased by 31.1% from RMB 292.7 million in 2022 to RMB 383.7 million in 2023, with a notable rise in ADAS-related revenue[39] Operational Efficiency and Cost Management - Total operating expenses for R&D, management, and sales decreased by 10.6% to RMB 5,795.4 million[11] - Capital expenditures, excluding one-time office-related investments, were reduced by 20%[11] - R&D expenses decreased by 13.7% from RMB 4,014.3 million in 2022 to RMB 3,465.8 million in 2023, attributed to reduced employee benefits and lower server operation costs[42] Generative AI Growth - In 2023, SenseTime achieved over RMB 1 billion in revenue from its generative AI business, marking the fastest growth of a new business since its establishment ten years ago[12][14] - Over 70% of customers in the generative AI segment were new clients in the past 12 months, with existing customers seeing an average increase of approximately 50% in their spending[14] - SenseTime's generative AI capabilities have seen a nearly 120-fold increase in usage for consumer-facing applications[18] - The traditional AI business's revenue share decreased from 81.9% in 2022 to 53.9% in 2023, reflecting a strategic shift towards generative AI technologies[25] Infrastructure and Technology Development - The company expanded its operational computing power to 12,000P and deployed 45,000 GPUs across multiple cities, enhancing its infrastructure for large model training services[15] - The cost-effectiveness of inference services improved by three times within a year, providing clients with the most competitive large-scale elastic inference services in the industry[15] - SenseTime's "Riri Xin" model series is undergoing continuous upgrades, with the upcoming version 5.0 expected to match GPT-4-turbo standards and support lossless context of up to one million words[20] - The "日日新" model versions 2.0 and 3.0 were released in July and August 2023, surpassing GPT-3.5-turbo performance, with the 4.0 version achieving capabilities comparable to GPT-4 in various applications[22] Corporate Governance and Compliance - The board is committed to high standards of corporate governance and has adhered to applicable code provisions throughout the reporting period[80] - The company has established a corporate governance committee to ensure operations align with shareholder interests and compliance with listing rules[104] - The audit committee has reviewed the group’s financial reporting procedures and internal control systems, ensuring compliance with applicable accounting standards[100] - The company has mechanisms in place to maintain board independence, including the majority of board committees being composed of independent non-executive directors[84] Shareholder Communication and Relations - The company emphasizes effective communication with shareholders to enhance investor relations and transparency[134] - The company has implemented a shareholder communication policy to ensure timely and equal access to information for shareholders[140] - The company encourages shareholders to provide their email addresses for timely communication[137] - The company’s website serves as a platform for communication with shareholders and investors, regularly updating information[138] Employee and Talent Management - The company emphasizes a culture of accountability, focusing on employee development, workplace safety, and sustainability, which helps attract and retain top talent[82] - The gender ratio among employees is 74.3% male and 25.7% female, with a goal to maintain at least the current level of gender diversity[112] - The company is committed to promoting gender diversity across its workforce, despite challenges in hiring skilled professionals in the AI industry[112] Risk Management - The company has established a comprehensive approach to operational risk management, focusing on identifying, measuring, monitoring, and controlling risks to minimize potential losses[121] - Compliance risk management is a core aspect of the company's risk management activities, ensuring operations comply with applicable laws and regulations[122] - The company has implemented standard protective measures for data security, including confidentiality classification, access control, and data encryption[124] Future Plans and Strategic Direction - The company plans to accelerate the commercialization of generative AI solutions, focusing on providing cost-effective options to users while promoting steady growth in traditional AI businesses[35] - The company aims to leverage the synergy of large devices and large models to enhance its service capabilities and maintain industry-leading model performance through continuous investment in model iteration[35] - The company plans to launch urban NOA functionality in 2024, continuing to lead the domestic market in autonomous driving technology[27]

Investment Rating - The report maintains a "Buy - A" rating for the company [3][6] Core Views - The company has achieved a comprehensive "cloud-edge" AI layout, enhancing its product capabilities with the launch of the new "Riri Xin Model 5.0" [1] - The company has a strong market competitiveness due to its unique AI stack capabilities and extensive experience in serving top clients across various sectors [1] - The generative AI business has seen rapid revenue growth, with a 199.9% year-on-year increase in 2023, contributing to 34.8% of total revenue [1][2] Summary by Sections Company Overview - The company held a technology exchange day on April 23, 2024, unveiling the upgraded "Riri Xin Model 5.0" [1] - The new model features a hybrid expert architecture with a context window of approximately 200K, matching the capabilities of GPT-4 turbo [1] Market Position - The company is recognized as one of the most authentic AI firms in China, with a controllable computing power of 12,000P and 45,000 GPUs [1] - It has successfully empowered clients in government, finance, healthcare, internet, and automotive sectors, showcasing its engineering capabilities [1] Financial Performance - In 2023, the company's generative AI revenue reached 1.184 billion, up 199.9% from the previous year, with revenue share increasing from 10.4% in 2022 to 34.8% [1][2] - The company expects to achieve revenues of 4.521 billion, 5.710 billion, and 7.077 billion for the years 2024, 2025, and 2026 respectively [2] Investment Outlook - The report suggests that the company is well-positioned to capture market share due to its technological capabilities and customer reputation, leading to anticipated revenue growth [2] - A target price of 1.02 HKD is set for the next six months, reflecting a 7x price-to-sales ratio for 2024 [2][3]

计算机 | 证券研究报告 — 调整盈利预测 2024年4月2日 00020.HK 商汤-W 买入 生成式AI收入高速增长，日日新5.0即将上线 原评级：买入 市场价格:港币0.71 根据公司 2023 年业绩公告，商汤 2023 年营业收入 34.1 亿元，同比-10.57%； GAAP归母净亏损-64.4亿元，同比扩大 6.54%；经调整亏损净额-54.1亿元，同 板 块评级:强于大市 比扩大 14.31%。我们预期公司将继续聚焦生成式 AI 核心业务，保持算力、算 股价表现 法、大模型领域优势，持续推动业务增长，有效管理开支，推动核心业务盈 利。维持买入评级。 25% 支撑评级的要点 5%  生成式 AI收入高速增长，日日新 5.0即将上线。在 2023年的业务重组中， (14%) 商汤业务板块改组为生成式 AI、传统 AI和智能汽车三大板块，后续生成式 (34%) AI 业务将成为公司重点发展方向，该业务主要商业模式有公有云模式、私 (54%) 有化部署模式以及定制化开发模式。2023生成式 AI业务收入 11.8亿元，同 (74%) 比+199.9%。主要来源于市场对生成式 AI 模型在训练、微调以 ...

Core Insights - SenseTime has experienced explosive growth in its generative AI business, achieving total revenue of 3.4 billion yuan in 2023, with the generative AI segment contributing 1.18 billion yuan, marking a nearly 200% growth rate and accounting for 35% of total revenue [1] - The company positions itself at the forefront of the AI 2.0 era, which signifies a shift from rule-based automation to deep learning and autonomous learning capabilities, fundamentally transforming AI applications across various sectors [2][4] - SenseTime's "SenseCore" infrastructure has become a critical asset, supporting the rapid development and deployment of large models, with a total computing power exceeding 12,000 petaFLOPS and 45,000 GPUs [3][6] Business Strategy - The integration of "large models + large infrastructure" has enhanced SenseTime's training capabilities, achieving over 90% acceleration efficiency and stable training periods exceeding 30 days [4][5] - The company has established partnerships with major players across various industries, including Qualcomm, major telecom operators, and leading financial institutions, which has broadened its market reach [4][5] - SenseTime has introduced a "model as a service" business model, allowing clients to easily fine-tune and utilize generative AI functionalities without the burden of managing underlying infrastructure [3][4] Market Position - Over 70% of SenseTime's generative AI clients in the past year were new customers, indicating strong market recognition and demand for its technology [5] - The company has developed high-performance edge models that outperform competitors like Meta's Llama2 and Google's Gemma, addressing the market's need for efficient and cost-effective AI solutions [5][6] - SenseTime's generative AI tools, such as "Xiaohuanxiong" family, have demonstrated superior performance in practical applications, enhancing productivity in programming and data analysis [7] Future Outlook - The company anticipates continued rapid growth in its generative AI business in 2024, driven by its technological advancements and successful integration of AI in various sectors [7] - SenseTime is positioned as a key player in the AI industry, driving upgrades and transformations while redefining productivity tools through innovative generative AI solutions [6][7]

Core Viewpoint - SenseTime has implemented a large device + large model industrial layout in 2023, significantly enhancing its computational power and model capabilities, positioning itself as a leader in generative AI in China [1][2]. Computational Power and Model Capabilities - The total operational computational power of SenseTime's large devices has surpassed 12,000 Petaflops, doubling since the beginning of 2023, with 45,000 GPUs deployed [1]. - The company's "Riri New" large model has achieved advanced training capabilities, including multi-modal and programming functionalities, establishing a leading position domestically [1][2]. Business Segments and Revenue Growth - Generative AI business revenue reached 1.18 billion RMB in the previous year, a 200% year-on-year increase, accounting for 35% of total revenue [1]. - Traditional AI business revenue was 1.84 billion RMB, down 41% year-on-year, with its revenue share declining from 82% in 2022 to 54% in 2023 [1]. - The automotive segment, represented by "Jueying," generated 380 million RMB in revenue, up 31% year-on-year, contributing 11% to total revenue [1]. Strategic Goals for 2024 - The company aims to maintain technological leadership, accelerate the commercialization of generative AI, and improve operational efficiency by focusing resources on core business areas [2][3]. Market Trends and Future Outlook - The demand for generative AI is driven by widespread needs across various industries, indicating the start of a new investment cycle in China's hard technology sector [1][2]. - The company anticipates continued rapid growth in generative AI, projecting that this segment could account for approximately 50% of total revenue by 2024, with a compound annual growth rate (CAGR) of around 50% from 2023 to 2028 [5][6]. Chip Utilization and Market Opportunities - The demand for computational power is expected to grow exponentially, with predictions of a tenfold increase in demand annually, creating opportunities for domestic chip development and optimization [3][4]. - The differentiation in chip usage between training and inference is anticipated to provide new design opportunities for chips tailored to specific tasks [4]. Business Models and Financial Projections - The generative AI business operates under three main models: public cloud, private deployment, and customized development services, with various pricing strategies based on usage [5]. - The financial impact of generative AI on gross margins is expected to be neutral, with current gross margins ranging between 40% and 45% [5]. Governance and Strategic Stability - Following the passing of a key figure, the ownership has been transferred to family members who continue to support the management team, ensuring stability in governance and strategic direction [6]. - The company aims to leverage its historical strengths to drive transformation in the generative AI era while maintaining healthy cash flow from existing operations [6].

证券研究报告 公 2024年03月28日 司 报 告 商汤-W(00020) 行 业： 计算机/IT服务Ⅱ │ 港 投资评级： 股 当前价格： 0.70港元 业务承压，全面加码生成式 AI - 公 目标价格： 司 事件：2023年3月26日，公司发布2023年财报，2023年实现营收34.06 年 基本数据 报 亿元（YoY-10.6%）,毛利约为15亿元（YoY-41%），毛利率为44.1%（YoY- 33,468.93/33,468. 点 总股本/流通股本(百万股) 22.7pct），年度亏损约为64.95 亿元（亏损增加 6.6%），EBITDA 亏损收窄 93 评 流通市值(百万港元) 23,428.25 至54亿元（亏损收窄2.2%）。 每股净资产(元) 0.86  生成式AI业务高速增长，传统AI业务全面收缩。2023年公司重新明 资产负债率(%) 25.01 确了生成式 AI、传统 AI 业务以及智能汽车三大业务板块。其中，生成式 一年内最高/最低(港元) 3.70/0.70 AI收入达11.84亿元，同比增速高达200%，占集团收入的比例提升24.4pct 至34.8%。主要得益于公司在从底 ...

Core Insights - SenseTime Group has established a clear business positioning with three core segments: Generative AI, Traditional AI, and Intelligent Vehicles [1] - The Generative AI segment has been designated as a strategic priority, becoming the main development direction for the company [1] - The company achieved remarkable growth in its Generative AI business, with revenue reaching 1.2 billion yuan, marking the fastest revenue growth for a new business in its ten-year history [1] Business Performance - The Generative AI segment's revenue growth was explosive, achieving a year-on-year increase of 200% and accounting for over 35% of total revenue [2][1] - The total computing power of SenseTime's large model infrastructure, SenseCore, has significantly increased to 12,000 petaFLOPS, with 45,000 operational GPUs [1] Innovation and Market Strategy - SenseTime has introduced a new "Model as a Service" business model, allowing clients to easily fine-tune and access various generative AI capabilities without the need to manage underlying infrastructure [1] - This innovative business model enhances customer experience and expands the company's market channels and business scope [1]

业绩简评 2024 年 3 月 26日，公司发布 2023 年报。公司2023年全年实现 营收 34.1 亿元，同比下滑 10.6%；毛利率为 44.1%，同比下滑 22.7pct，毛利润为 15.0 亿元，同比下滑 41.0%；净利润为-64.9 亿元，亏损扩大6.6%；经调整EBITDA为-43.7 亿元，亏损扩大 3.7%。 2023 年下半年，公司营收为 19.7 亿元，同比下滑 17.6%；毛利 率为 43.2%，同比下滑 24.0pct，毛利润为 8.5 亿元，同比下滑 47.0%；净利润为-33.5 亿元，亏损扩大16.2%；经调整EBITDA 为 23.4 亿元，亏损扩大23.9%。 经营分析 分业务而言：1）生成式AI业务 2023 年实现营收11.8 亿元，同 比增长199.9%，主要受益于公司大装置及大模型服务能力的领 先。公司目前总算力达到1.2 万 petaFLOPS，运营GPU数量达 4.5 万张。2）智能汽车业务2023 年实现营收3.8 亿元，同比增 港币(元) 成交金额(百万元) 4.00 7,000 长 31.1%；2023 年下半年智能汽车营收为3.0亿元，同比增长 6, ...

Financial Performance - Total revenue for the year ended December 31, 2023, was RMB 3,405.8 million, a decrease of 10.6% compared to RMB 3,808.5 million in 2022[5]. - Gross profit for the same period was RMB 1,500.8 million, with a gross margin of 44.1%, down from 66.8% in 2022[5]. - The company reported a net loss of RMB 6,494.7 million, an increase in loss of 6.6% from RMB 6,092.99 million in the previous year[5]. - Revenue for the year ended December 31, 2023, was RMB 3,405,842 thousand, a decrease of 10.6% compared to RMB 3,808,510 thousand in 2022[39]. - Gross profit for 2023 was RMB 1,500,787 thousand, down 41.0% from RMB 2,542,267 thousand in 2022[39]. - Operating loss for 2023 was RMB 6,646,555 thousand, slightly increased from RMB 6,528,323 thousand in 2022[39]. - The annual loss for 2023 was RMB 6,494.7 million, compared to a loss of RMB 6,093.0 million in 2022, as operational cost savings were insufficient to offset the decline in gross profit[54]. - The total comprehensive loss for the year amounted to RMB 6,280.361 million, up from RMB 4,147.653 million in 2022[84]. Revenue Breakdown - Revenue from generative AI reached RMB 1,183.7 million, representing a rapid growth of 199.9% and accounting for 34.8% of total revenue[7]. - Revenue from the intelligent vehicle business was RMB 383.7 million, a year-on-year increase of 31.1%, contributing 11.3% to total revenue[7]. - Traditional AI business revenue declined by 41.1% to RMB 1,838.4 million, reducing its share of total revenue from 81.9% in 2022 to 53.9% in 2023[7]. - Revenue decreased by 10.6% from RMB 3,808.5 million in 2022 to RMB 3,405.8 million in 2023, primarily due to a decline in traditional AI revenue, partially offset by increases in generative AI and smart vehicle revenues[40]. - Generative AI revenue surged by 199.9%, rising from RMB 394.7 million in 2022 to RMB 1,183.7 million in 2023, driven by explosive demand for generative AI model training, fine-tuning, and reasoning[40]. - Traditional AI revenue fell by 41.1% from RMB 3,121.1 million in 2022 to RMB 1,838.4 million in 2023, reflecting a significant reduction in reliance on smart city business[41]. - Smart vehicle revenue increased by 31.1%, from RMB 292.7 million in 2022 to RMB 383.7 million in 2023, with a notable rise in ADAS-related revenue[41]. Operational Metrics - The company’s EBITDA loss was RMB 5,449.6 million, a decrease of 2.2% from RMB 5,570.2 million in 2022[5]. - Trade receivables increased by 48.5% to RMB 3,885 million[7]. - The total computing power reached 12,000 petaFLOPS, with 45,000 GPUs operational[7]. - The company has seen a significant increase in strategic customer collaborations, with the number of clients with orders exceeding 10 million RMB reaching several dozen[16]. - The company has established partnerships with major smartphone brands like Xiaomi and Honor to develop innovative features for AI applications[25]. Research and Development - The company is committed to the development of version 5.0 of the "Daily New" model, expected to be released in April 2024, which will fully meet the standards of GPT-4-turbo[20]. - The company has established an advanced end small model that provides performance comparable to large models but is small enough to run on laptops or mobile devices[15]. - The company continues to invest in research and development, with R&D expenses amounting to RMB 503,596,000 in 2023, up from RMB 458,755,000 in 2022, marking an increase of about 9.8%[151]. - The company has engaged an independent external appraiser to assess the recoverable amounts of cash-generating units, ensuring the accuracy and reliability of financial information[153]. Cash Flow and Liquidity - The total cash and cash equivalents increased to RMB 9,423.5 million as of December 31, 2023, up from RMB 7,962.8 million at the end of 2022, marking an increase of about 18.3%[61]. - The net cash used in operating activities for 2023 was RMB (3,234.3) million, compared to RMB (3,084.5) million in 2022, indicating a worsening of approximately 4.9%[62]. - The cash flow from financing activities generated RMB 1,083.6 million in 2023, a decrease from RMB 3,329.1 million in 2022, indicating a decline of about 67.5%[62]. - The company primarily met its cash needs through shareholder contributions, with cash and cash equivalents significantly increasing over the year[61]. Assets and Liabilities - Total assets as of December 31, 2023, amounted to RMB 32,888,024 thousand, a decrease from RMB 37,426,993 thousand in 2022[86]. - Total liabilities rose to RMB 9,732,551 thousand in 2023, compared to RMB 8,413,133 thousand in 2022, indicating an increase of about 15.6%[88]. - The company reported a net cash inflow from financial assets of RMB 12,697 thousand in 2023, down from RMB 18,307 thousand in 2022[94]. - The company has pledged assets with a book value of RMB 4,336.0 million in properties and land use rights, along with RMB 324.0 million in restricted cash to secure loans[80]. Future Outlook - The company plans to accelerate the commercialization of generative AI and enhance market penetration, aiming to provide cost-effective solutions[36]. - The focus for 2024 includes leveraging the synergy of large devices and models to enhance service capabilities and reduce reasoning costs[36]. - The company plans to achieve mass production of urban NOA functionality in 2024, aiming to maintain a leading position in the domestic market for autonomous driving deliveries[27]. - The company anticipates continued enhancement of model capabilities, with expectations for significant advancements in generative AI applications in enterprises[37]. Governance and Compliance - The company is committed to maintaining a "zero tolerance" policy towards fraud and corruption, ensuring a transparent business environment[35]. - The company received multiple certifications, including ISO 27001 and ISO 27701, demonstrating its commitment to information security and privacy management[34]. - The board has decided not to declare a final dividend for the year ending December 31, 2023[197].

Financial Data and Key Metrics Changes - Total annual revenue for the group in 2023 was RMB 3,400,000,000, a decrease of 11% compared to the previous year, primarily due to a strategic reduction in reliance on the smart city business [11][49] - EBITDA losses narrowed to RMB 5,500,000,000, with operating expenses in R&D, G&A, and sales decreasing by 10.6% year on year [12][50] - Trade receivables collected in 2023 amounted to RMB 3,800,000,000, reflecting a year-on-year growth of 48.5% [11][54] Business Line Data and Key Metrics Changes - The generative AI business achieved revenue of RMB 1,180,000,000, representing a 200% year-on-year increase and accounting for 35% of total revenue [15][49] - Traditional AI revenue decreased by 41%, generating RMB 1,840,000,000 [15] - The smart auto business generated RMB 384,000,000 in revenue, a 31% increase, accounting for 11% of total revenue [16] Market Data and Key Metrics Changes - The smart city segment's revenue contribution significantly declined from about 30% in 2022 to less than 10% in 2023 [11][39] - The generative AI business has become the core business, expected to continue rapid growth in 2024 [16][17] Company Strategy and Development Direction - The company is focusing on generative AI, traditional AI, and smart auto as its three core business segments, with a strong emphasis on developing the generative AI business [14][15] - The company aims to optimize operational efficiency and reduce losses by incubating and spinning off non-core businesses [12][47] - Future strategic goals include maintaining technological leadership, driving business growth, and promoting the core business [46][47] Management's Comments on Operating Environment and Future Outlook - Management highlighted the rapid advancements in AI technology and the significant demand for generative AI across various industries [8][17] - The company anticipates that generative AI will account for about half of total revenue in 2024, with a projected CAGR of around 50% from 2023 to 2028 [73] - Management expressed confidence in overcoming challenges during the transformation period and securing a leading position in the AI industry [48][80] Other Important Information - The company underwent business restructuring in 2023, focusing on generative AI and reducing reliance on the smart city business [11][14] - The passing of the founder, Professor Tang, was acknowledged, with management committing to continue his legacy in AI innovation [6][7] Q&A Session Summary Question: What is SenseTime's exploration in the multimodal domain? - Management emphasized the importance of historical technology accumulation and the role of synthetic data in multimodal research, predicting its significance in future developments [60] Question: What are the expansion plans for computing power for 2024 and beyond? - The company plans to double its computing power in 2024, addressing the increasing demand for both training and inference in large models [63][64] Question: What are the current business models for generative AI and growth expectations? - The company operates three business models: public cloud, private cloud, and customized model services, with expectations for generative AI to contribute significantly to total revenue in the coming years [71][73] Question: What are the impacts of the founder's passing on shareholding and strategy? - The founder's shares were inherited by his family, who continue to support the management, ensuring stability in governance and strategy [75][80]