Core Insights - The year 2025 is anticipated to be a pivotal year for the implementation of large models in AI, with companies needing to overcome three major barriers: dynamic balance of heterogeneous computing power supply, controllability of model iteration, and integration of technology with business [1] - Xinhua San's senior vice president highlighted the rapid acceleration of the AI industry, indicating a shift from merely observing AI developments to understanding its practical applications [1] Infrastructure Development Trends - The development of AI infrastructure is focusing on two core trends: ubiquitous collaboration of diverse computing power and integration of cloud, edge, and terminal [3] - Xinhua San is adapting to diverse computing needs by offering a wide range of products, from servers to supernodes, to enhance computing efficiency [3] - High bandwidth, low latency, and high reliability in network transmission are critical, with Ethernet technology showing significant advantages [3] Open Ecosystem Construction - Xinhua San maintains an open approach to ecosystem development, having adapted over 80 GPU cards from major domestic manufacturers [4] - The architecture of Xinhua San's supernodes is designed in collaboration with GPU manufacturers to define technical paths from the outset [4] - The company is innovating its software platform and business models to enhance customer capabilities in AI applications [4] Supernode Value Proposition - The core value of Xinhua San's supernodes lies in improving communication efficiency between cards, which enhances training and inference performance [5] - The supernodes are particularly advantageous in large-scale training scenarios, supporting various AI applications [5] - The company emphasizes the importance of an open and standardized ecosystem to foster industry growth [5] Balancing Short-term and Long-term Needs - Xinhua San is addressing the balance between immediate demands and long-term computing power planning, particularly for small and medium enterprises [6] - The introduction of all-in-one machines significantly reduces deployment time and lowers barriers for AI adoption [6] Future Outlook - The future of AI will focus on the collaboration of computing, storage, and networking, with Xinhua San enhancing its offerings in these areas [7] - Industries such as education, research, finance, and electricity are expected to drive significant demand for computing power [7] - Xinhua San aims to provide an open architecture that accommodates diverse computing power while ensuring efficient delivery and reduced switching costs [7]