en.Wedoany.com Reported - On June 25, Hygon Information Technology Co., Ltd. and Tongji University officially signed a strategic cooperation agreement to jointly launch the nation's first domestically produced thousand-card engineering AI computing platform. Powered by the domestic Hygon DCU, the platform adopts a super-fusion architecture capable of simultaneously supporting high-performance computing and AI training and inference tasks. It is primarily designed for real-world industrial scenarios such as engineering simulation and intelligent construction, and has been integrated into Tongji University's campus-level computing service system for operational use.
The significance of this platform goes beyond the "thousand-card scale"; it directly places domestic computing power into engineering research and application scenarios. In the past, universities built AI computing platforms primarily for general AI training, large model inference, or scientific experiments. This platform, however, is explicitly oriented towards AI4E (AI for Engineering), emphasizing dedicated computing support for engineering simulation, complex structural calculations, intelligent construction, urban infrastructure, transportation systems, and manufacturing systems.
The Hygon DCU serves as the core computing unit of the platform. Designed for high-performance computing, artificial intelligence, and large-scale parallel computing tasks, the DCU can support matrix calculations, finite element analysis, fluid dynamics computations, model training, and inference services in engineering simulation. For engineering scenarios, a computing platform must not only consider peak performance but also balance scientific computing, AI tasks, engineering software ecosystems, domestic hardware and software compatibility, and long-term stable operation.
The academic disciplines at Tongji University provide a clear application direction for the platform. Fields such as civil engineering, transportation, architecture and planning, intelligent construction, automotive engineering, and urban governance all require substantial high-performance computing and AI model support. Tasks like large-scale bridge structural simulation, urban traffic flow prediction, building lifecycle management, intelligent construction process optimization, and complex system safety assessment demand high levels of computing scale, scheduling efficiency, and data processing capability.
The platform adopts a super-fusion architecture, meaning traditional high-performance computing and AI computing are no longer separated into two systems. Engineering simulation typically relies on high-precision numerical computing, while AI training and inference depend more on large-scale parallel computing power and data-driven models. By running both types of tasks collaboratively on the same platform, researchers can complete model training, simulation validation, result analysis, and engineering optimization in a unified environment, reducing the costs associated with cross-platform migration.
Sha Chaoqun, President of Hygon Information Technology, stated that this cooperation marks a key milestone in the transition of domestic computing power from AI4S (AI for Science) to AI4E (AI for Engineering). While AI4S emphasizes scientific discovery and fundamental research, AI4E directly targets engineering design, construction, and industrial applications. For the domestic computing power industry, the ability to enter real engineering scenarios determines whether the platform's value remains confined to experimental validation or can genuinely serve industrial upgrading.
The launch of the nation's first domestically produced thousand-card engineering AI computing platform also provides a new model for university computing infrastructure development. It is not merely a procurement of computing cards but the formation of a platform-based capability centered on engineering education, engineering research, and industry collaboration. Students and research teams can leverage domestic computing power within the campus-level system for model training, simulation calculations, and engineering validation. Enterprises also have the opportunity to bring engineering problems into the university platform for joint research through industry-academia collaboration.
As AI moves from general-purpose models to industry-specific applications, computing platforms need to be more closely aligned with specific industrial scenarios. The engineering field, with its complex data structures, long simulation cycles, and high model validation requirements, imposes stronger demands on hardware-software synergy. The launch of the domestic thousand-card engineering AI computing platform by Hygon and Tongji University marks the entry of domestic computing power into a phase of large-scale validation in university engineering research and real-world industrial applications. Going forward, attention should be paid to the platform's actual performance in intelligent construction, engineering simulation, and other AI4E scenarios, as well as the progress of the domestic DCU ecosystem in adapting to mainstream engineering software and AI frameworks.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
