JD Cloud and Moore Threads are planning a 100,000-GPU intelligent computing cluster, marking a significant expansion in China’s effort to build large-scale AI infrastructure around domestically developed chips. The project brings together JD Cloud’s cloud-computing capabilities with Moore Threads’ Universal GPUs at a scale that could test whether China’s homegrown accelerator ecosystem can support demanding commercial workloads. Moore Threads described the initiative as the first deployment of domestically developed GPUs in a 100,000-card core computing cluster at a leading Chinese AI cloud provider. The company views the project as a move from smaller deployments toward wider commercial adoption of domestic intelligent-computing technology.
Universal GPUs Target Large-Scale AI Workloads
The planned cluster will use Moore Threads’ Universal GPUs to handle large-model training and inference, embodied intelligence, and other computing-intensive applications. Its capacity will also serve companies across industries, creating a shared computing resource rather than a system limited to a single AI development program. That model reflects the growing importance of infrastructure that can support multiple workloads as companies seek more computing power for increasingly sophisticated AI systems. The economics of such clusters also depend on sustained utilization, making access to external enterprise workloads an important part of the project’s commercial logic.
JD Cloud plans to direct additional computing resources toward supply-chain AI, embodied intelligence, integrated model training and inference, and industrial applications. Demand for these workloads is pushing Chinese technology companies toward GPUs capable of coordinating large training jobs across thousands or tens of thousands of processors. Embodied intelligence adds another layer of pressure because AI systems that interact with physical environments require substantial resources for training, simulation, inference, and data generation. As these applications move closer to commercial deployment, computing infrastructure becomes a strategic component of the broader AI product stack.
Moore Threads Builds Around Domestic GPU Scale
Moore Threads has developed GPUs designed for large-scale AI training and has deployed clusters ranging from thousands to tens of thousands of GPUs on a single network. According to the company, those systems have supported foundation models, embodied-intelligence models, and world models, giving the chip developer experience with workloads that extend beyond conventional AI inference. The 100,000-GPU target represents a substantially larger infrastructure ambition and could provide a high-profile test of how domestic accelerators perform when deployed at greater scale. For Moore Threads, the project also offers an opportunity to demonstrate that its hardware can participate in increasingly integrated AI computing environments.
The partnership will cover the technology stack from chips and cloud platforms through model training, according to Moore Threads. The companies intend to support continued development of JD’s JoyAI models while accelerating model training, simulation, and high-quality data generation. Their stated objective is to create an integrated pipeline spanning data, training, simulation, and deployment, connecting computing infrastructure more directly with AI development. That approach could reduce fragmentation between hardware resources and model-development workflows as Chinese companies build increasingly complex AI systems.
Beijing Signals Demand for Massive Clusters
The project arrives as Chinese policymakers place greater emphasis on large intelligent-computing facilities and domestic semiconductor technology. On Monday, China’s Ministry of Industry and Information Technology released a plan for the information and communications industry during the 15th Five-Year Plan period. The plan calls for the orderly deployment of intelligent-computing clusters equipped with 10,000, 100,000 or more accelerator cards, along with inference-computing facilities tailored to specific application needs. It also calls for stronger efforts to adapt infrastructure to domestically developed computing chips, closely matching the direction of the JD Cloud and Moore Threads project.
Founded in 2020, Moore Threads develops GPUs for artificial intelligence and other high-performance computing workloads. The company listed on Shanghai’s STAR Market in December 2025, becoming one of China’s most prominent publicly traded GPU developers. Its expansion comes as Chinese technology companies face pressure to develop alternatives to foreign computing hardware while maintaining access to the processing capacity required for advanced AI. Meanwhile, a 100,000-GPU deployment could give domestic suppliers a valuable real-world environment for demonstrating scalability across cloud infrastructure and enterprise applications.
Scale Becomes the Next AI Infrastructure Test
The significance of the project extends beyond the headline number of GPUs because large clusters require coordination across hardware, networking, cloud software, models, and data pipelines. A successful deployment could strengthen the case for domestic accelerators in commercial AI environments where training and inference workloads increasingly operate at massive scale. Moreover, broader access to the cluster could allow businesses in manufacturing, supply chains, robotics, and other industries to tap computing capacity without building comparable infrastructure independently. The partnership therefore positions computing scale, rather than individual chip performance alone, as an increasingly important measure of China’s progress in developing an independent AI infrastructure ecosystem.
