JD Cloud announced on September 9, 2026, during the 2026 Global Technology Explorers Conference, an operational initiative to construct a computing cluster powered by 100,000 general-purpose graphics processing units manufactured by Chinese semiconductor producer Moore Threads. The planned computing center will dedicate its hardware resources to processing massive model training cycles and executing real-time inference workloads, while directing specialized technical capacity toward embodied artificial intelligence projects. Processing capacity generated by this distributed network will be made commercially available to corporate clients operating across diverse industry sectors.

Allocation of computing capacity across commercial sectors

Moore Threads supplies the hardware.

Under the framework outlined at the 2026 Global Technology Explorers Conference, JD Cloud designed this massive infrastructure to sustain concurrent large-model training workloads and continuous real-time algorithmic inference operations while distributing dedicated computing capacity directly to commercial enterprises operating across a wide range of industrial markets. Organizations utilizing the facility will access centralized processing capacity through standard cloud service interfaces, removing the requirement for independent firms to purchase and maintain their own physical server arrays. The computing network separates computational power into scalable allocations, allowing commercial businesses to train proprietary machine learning models or execute immediate data analysis depending on their specific operational demands. Enterprise clients can direct these computing resources toward digital services, financial analytics, and customer interaction platforms. This infrastructure model guarantees that enterprises of varying sizes can deploy intensive computational algorithms across commercial software environments.

Development of physical artificial intelligence applications

JD Cloud stated that building this large-scale computing infrastructure based on domestic hardware will help bring physical artificial intelligence into practical commercial use, supporting the integration of complex machine learning models directly into physical machines and autonomous equipment. Developing functional systems for embodied artificial intelligence demands substantial computing density to process sensory information, calibrate physical interactions, and coordinate automated mechanical movement in real-world industrial environments.

Milestone for domestic graphics processing units

JD Cloud described the project as the first deployment of Chinese graphics processing units within a 100,000-GPU core computing cluster at a major domestic cloud provider. Moore Threads developed these general-purpose processors to handle diverse computational tasks, providing an alternative processing architecture capable of executing high-intensity matrix mathematics and complex algorithmic functions. By demonstrating that 100,000 domestic units can function in a single coordinated cluster, the initiative confirms the technical viability of deploying homegrown processors within primary enterprise data centers. The infrastructure demonstrates that domestic hardware components can fulfill large-scale operational requirements for critical commercial computing facilities.

JD Cloud unveiled the project roadmap during technical presentations at the 2026 Global Technology Explorers Conference.