对话云天励飞董事长陈宁：预计到2030年，每台设备都将内置AI推理芯片

Core Viewpoint - The article discusses the strategic shift of Yuntian Lifei towards AI inference chips, highlighting the growing demand for inference computing power as AI applications proliferate, and the company's recent product launches aimed at this market [2][3][4]. Group 1: AI Inference Market Dynamics - AI is divided into training and inference stages, with inference power demand increasing significantly; IDC reports that cloud inference computing power reached 58.5% in 2022 and is expected to rise to 62.2% by 2026 [2]. - The AI inference chip market is still in a blue ocean phase, with growth expected to outpace the training chip market in the next three to five years, indicating a transition from the training era to the inference era [3][4]. - The demand for AI inference computing is projected to grow over 80% annually, potentially surpassing training as the main driver of data center computing [15]. Group 2: Company Developments and Product Launches - Yuntian Lifei launched the DeepSky X6000 Mesh inference computing card with a computing power of 256 Tops, designed for AI inference applications, supporting 30fps 1080p video decoding [2]. - The company has developed a range of products based on the DeepSky X6000 Mesh card, including various integrated machines with up to 4 PFlops of inference computing power [2][4]. - Yuntian Lifei has successfully transitioned to domestic processes, becoming the first AI chip company to adopt domestic Chiplet technology, and has launched multiple chips under the DeepEdge10 platform [6]. Group 3: Financial Performance and Future Outlook - Yuntian Lifei reported a revenue of over 900 million yuan in 2024, a year-on-year increase of 81.3%, with Q1 2025 revenue reaching 264 million yuan, up 168.23% year-on-year [7]. - The company has signed a significant contract for AI training and inference heterogeneous computing services, expected to contribute approximately 1.6 billion yuan in revenue over three years [7]. - The CEO emphasizes the importance of AI inference chips in driving the large-scale application of AI technologies in various sectors, predicting a transformative impact on electronic products in the next five years [8][9].