JD Cloud to Build 100,000-GPU AI Cluster with Moore Threads

On September 9, 2026, at the JD Global Tech Explorer Conference in Beijing, JD Cloud unveiled plans to construct a massive AI computing cluster featuring 100,000 GPUs. The cluster will leverage Moore Threads' full-function GPUs as its computational foundation, targeting key areas such as large-scale model training, inference, and embodied intelligence. This initiative aims to open up computing resources across various industries, accelerating the industrialization and scalability of physical AI.

The announcement signifies that domestic GPUs have entered the tier of top-tier AI cloud enterprises for the first time, achieving a scale of 100,000 cards in core computing clusters. This transition marks the evolution of domestic intelligent computing from a state of 'usable' to 'commercially viable on a large scale'. The project represents a strategic step forward in building a national computing infrastructure with indigenous technology.