Sugon(SH:603019) Xin Lang Cai Jing·2026-02-06 11:20Core Insights - The article highlights the successful deployment of three sets of scaleX supercomputing clusters by Zhongke Shuguang, marking the first domestic implementation of over 30,000 cards in an AI computing pool in China [1][12]. Group 1: Deployment and Significance - The scaleX supercomputing cluster was first showcased at the HAIC conference in December last year and became operational within two months, demonstrating rapid advancement in domestic AI computing capabilities [3][14]. - This deployment signifies a critical transition from merely constructing infrastructure to actively utilizing it, representing a substantial leap for domestic supercomputing clusters [4][14]. Group 2: Demand and Evolution - The demand for computing power has surged as AI model parameters have escalated from billions to trillions, necessitating a shift in architecture to meet these requirements [5][15]. - Industry forecasts predict that by 2025, China's demand for intelligent computing power will increase nearly 20 times compared to 2020, with sustained high growth expected in the following years [6][16]. Group 3: Technological Advancements - The scaleX supercomputing cluster not only achieves a significant scale increase but also makes systematic breakthroughs in system architecture, energy efficiency management, and resource scheduling, pushing domestic AI computing into a new phase of quality and quantity [6][16]. - The cluster is designed based on an open architecture that fully supports mainstream software ecosystems like CUDA, facilitating mixed deployment of various domestic acceleration cards and lowering the adaptation threshold for developers [10][19]. Group 4: Application and Impact - The scaleX supercomputing cluster provides deep empowerment across multiple scenarios, including support for trillion-parameter model training, high-throughput inference for major internet companies, and advancements in AI for scientific research [11][20]. - It significantly enhances research efficiency in fields such as materials science, achieving improvements in protein research efficiency by 3 to 6 orders of magnitude [11][20]. Group 5: Future Outlook - The ongoing construction of the national supercomputing internet and the integration of computing power scheduling mechanisms are gradually establishing a self-controlled, open, and efficient AI computing system in China [12][20]. - The successful deployment of the scaleX supercomputing clusters is expected to leverage national-level computing hubs to provide high-quality AI computing services globally [12][20].