JD Cloud, the cloud computing arm of JD.com, said on September 9 at the 2026 JD Global Technology Explorers Conference that it plans to build a 100,000-GPU full-function cluster using chips from Moore Threads, a China-based GPU developer.

JD Cloud said the cluster will support large-model training, inference, and embodied intelligence. It added that the computing capacity will be made available to companies across the industry.

The company described the project as ultra-large-scale domestic intelligent computing infrastructure, but did not disclose a timeline, investment size, or rollout details.

JD Cloud said the project would mark the first use of domestic GPUs in a 100,000-card core intelligent computing cluster at a leading AI cloud company.

For context, JD.com’s infrastructure-as-a-service business ranks No. 4 in China, according to Gartner.

🔗 Source: Security Times

Recent JD Cloud developments