JD Cloud announced it will build a 100,000-card intelligent computing cluster based on full-featured GPUs from Moore Threads. This marks the first large-scale 100,000-card intelligent computing project deployed by a leading domestic cloud provider with domestic GPUs as the core computing hardware. Once completed, the cluster will offer computing rental services for large model training, AI inference, digital twins and industrial simulation, expanding domestic self-controlled AI computing supply and lowering Chinese enterprises’ reliance on overseas computing hardware.

For a long time, domestic AI computing supply heavily relied on high-end overseas GPUs. Overseas chip export restrictions have slowed capacity expansion for domestic large-model developers. The domestic GPU industrial chain keeps upgrading and gradually achieves commercial deployment in general computing, video encoding and multimodal model inference scenarios. A 100,000-card cluster counts as ultra-large computing infrastructure with strict requirements for chip stability, inter-node connection, heat dissipation and operation platforms. The project signals domestic GPUs are moving from small-scale pilots toward large commercial clusters.
Industry analysts mention intelligent computing clusters require heavy capital investment, long construction cycles and massive funding. The project proves domestic general-purpose GPUs have the basic capability to support large commercial computing projects and will drive order growth for upstream chip manufacturing, servers, high-speed interconnection and thermal management suppliers. Computing demand mainly originates from AI startups, traditional manufacturers undergoing digital transformation and research institutes. Risks remain as software ecosystems for domestic GPUs still have gaps, and continuous investment is needed for large model adaptation and optimisation. Market participants will watch the project launch timeline, growth in commercial orders for domestic GPUs and the total capacity of newly launched intelligent computing centres nationwide.