JD Cloud said it plans to build a 100,000-GPU intelligent-computing cluster powered by Moore Threads general-purpose GPUs, according to Chinese primary coverage of the 2026 JD Global Technology Explorers Conference on Sep. 9. Securities Times via Xinhua Finance and Sina used the wording “plans to build” a 100,000-card cluster. TechNode’s English report likewise framed a 100,000 general-purpose GPU plan from Moore Threads.
Coverage frames this as a plan, not a completed build. No capital expenditure figure was disclosed. Stated focus areas include large-model training, inference, and embodied or physical AI, with capacity described as opening to industry users.
The announcement is intent at headline scale for a domestic GPU vendor inside a major China cloud, with delivery and schedule risks unstated. Distinct from memory or HBM inventory stories.
Sources: Securities Times/Xinhua Finance via Sina, Sep. 9, 2026; TechNode.
