China's JD Cloud and chip maker Moore Threads are set to work together on a 100,000-strong GPU cluster.

Revealed at JD's 2026 Global Technology Explorers Conference, the project will use Chinese hardware. According to JD Cloud, this is the first cluster of its kind to be hosted by a major domestic AI cloud provider.

As reported by Chinese publication Securities Times, the cluster will comprise Moore Threads' "full-featured GPUs" and will be designed for large-scale model training, inference, and embodied intelligence workloads.

The two companies have previously partnered on a 10,000 GPU cluster.

Details about the GPUs have not been provided.

According to Moore Threads' website, the company has the MTT S5000 GPU, powered by the Moore Threads' PH100 chip and built on the advanced "PingHu" architecture. The company says it can deliver full-precision compute support from FP8 to FP64. The company also offers the MTT S4000 and S3000 GPUs, the latter of which is based on the ChunXiao chip.

Cloud provider JD Cloud operates data centers across China, with its core regions in Beijing, Guangzhou, Shanghai, and Suqian.

Moore Threads was founded by former Nvidia executive James Zhang Jianzhong in 2020. The company has received funding from Shenzhen Capital Group, GGV Capital, Sequoia Capital China, ByteDance, and Tencent, and went public on the Shanghai Stock Exchange in December 2025. The company is listed on the US Bureau of Industry and Security's Entity List.