Huaco AI server HK2498A2 series products have the characteristics of high expansion, high performance, high energy efficiency, flexible deployment, etc. AI computing performance can reach 5 trillion times per second, which is suitable for typical AI application scenarios such as image video, voice recognition, intelligent customer service, etc. With the massive growth of data and the acceleration of model update iteration, AI scientific research institutions and commercial companies urgently need to improve AI computing power to shorten the model training and development cycle. At the same time, they also hope to deploy AI infrastructure more quickly and economically, realize the compatibility of AI infrastructure with the original IT infrastructure, save data center space and reduce costs. The HK2498A2 adopts a high-speed interconnection technology interconnection structure, and the 4U space can carry 8 high-speed interconnection GPUs Cambricon MLU290-M5/MLU370-M8. Data P2P interaction can be directly carried out between any two GPU, and 448 GB/s transmission bandwidth can be realized. At the same time, two AMD EPYC 7000 series PCIe4.0CPU are configured to provide ultra-advanced general purpose computing performance in combination with XGMI-2 bus interconnection design. The 4U size and redundant power supply design make it suitable for a wider range of data center deployment environments, providing customers with extremely high reliability and stability; at the same time, it HK2498A2 the use of more flexible cluster deployment solutions, and can achieve integrated deployment from hardware to applications. In addition, the 54V power supply scheme is adopted in the system, which is more energy efficient. The design of hierarchical and regional cooling channels and intelligent control strategy can obtain the best cooling efficiency. HK2498A2 can help AI users efficiently complete the construction of AI infrastructure and development environment, while enjoying high computing performance while reducing deployment and operating costs.