Workflow
首次覆盖:生成式AI厚积薄发,AI龙头再出发
00020SENSETIME(00020) 国泰君安·2024-05-15 05:02

Investment Rating - The report initiates coverage with a "Buy" rating for the company [3][4]. Core Insights - SenseTime is a leading AI enterprise in China, focusing on generative AI, with a strategic shift in business focus to this area in 2023. The company aims to leverage its capabilities from computational infrastructure to large model deployment, anticipating significant growth driven by AI terminal and multimodal model demand [3][16]. - The generative AI market is expected to grow rapidly, with projections indicating a global market size of 150billionandaChinesemarketsizeofapproximately150 billion and a Chinese market size of approximately 13 billion by 2027, reflecting compound annual growth rates of 85.7% and 86.2%, respectively [3][23]. - The company has established a leading AI computational infrastructure, SenseCore, capable of supporting the entire lifecycle of trillion-parameter large models, which positions it well for future growth [3][25]. Summary by Sections 1. Earnings Forecast and Valuation - Revenue projections for SenseTime are estimated at 45.76 billion, 60.05 billion, and 77.06 billion RMB for 2024, 2025, and 2026, respectively. The net profit is expected to be -4.54 billion, -2.93 billion, and -1.44 billion RMB for the same years [10][14]. - The company is valued using a price-to-sales (P/S) ratio, with a target price set at 2.25 HKD based on a comparable company average P/S of 15.16 for 2024 [3][14]. 2. Generative AI Business Growth - The generative AI business has seen explosive growth, with revenue reaching 1.18 billion RMB in 2023, a 200% year-on-year increase, and now accounting for 35% of total revenue [17][21]. - Traditional AI business revenue has declined by 41% to 1.84 billion RMB, reflecting a strategic reduction in reliance on lower-quality cash flow segments like smart city projects [17][18]. 3. Technological Capabilities - SenseTime's SenseCore infrastructure supports large model training and has achieved over 90% acceleration efficiency in its training clusters, with a total operational computing power of 12,000 P and 45,000 GPUs [25][27]. - The company has rapidly iterated its "日日新" large model, achieving performance levels comparable to leading international models like GPT-4 Turbo [28][30]. 4. Market Potential - The generative AI sector is positioned as a key growth driver across various industries, with significant investments and developments occurring in finance, education, healthcare, and more [23][24]. - The report highlights the potential for generative AI to transform productivity and operational efficiency across sectors, indicating a broad market opportunity [21][23].