What happened
At the 2026 JD Global Technology Exploration Conference in Beijing's Yizhuang area on September 9, JD unveiled its 'Super AI Supply Chain' initiative, aiming to integrate computing power, data, models, hardware, and industry scenarios to scale AI applications in physical industries.
JD Cloud has entered deep cooperation with Moore Threads to build a 100,000-card domestic intelligent computing cluster based on full-function GPUs, serving large model training, inference, and embodied intelligence, with open computing services for the entire industry.
JD's technology committee chair and JD Cloud president Cao Peng stated that JD will continue to coordinate data, models, and supply chain capabilities around the strategic direction of building the world's largest physical world operation center, accelerating physical AI industrialization.
Why it matters
This move underscores JD's push to reduce reliance on foreign chips by leveraging domestic GPU technology, which is critical for AI sovereignty and supply chain resilience.
The 'Super AI Supply Chain' could set a precedent for how large enterprises integrate AI across physical operations, potentially influencing industry standards for data collection, model deployment, and hardware standardization.
The scale of the planned cluster and data collection efforts signals a major investment in physical AI, which could accelerate advancements in robotics and automation across retail, logistics, and other sectors.
Key facts
JD and Moore Threads plan to build a 100,000-card domestic intelligent computing cluster using full-function GPUs.
JD aims to collect 10 million hours of high-quality industrial data within two years, leveraging nearly 10,000 offline stores and 3,600 warehousing nodes.
Moore Threads reported first-half revenue of 1.736 billion yuan, a 147.42% year-on-year increase, with cloud product revenue accounting for about 97.5% of total revenue.
JD's AI technology has been deployed in over 3,000 real-world scenarios across retail, logistics, health, industry, food delivery, and housekeeping.
JD plans to support over 100 upstream and downstream manufacturers in the next three years and deploy over one million smart terminals in five years.
What to watch next
Monitor the progress of the 100,000-card cluster construction and whether Moore Threads can meet delivery targets, as this will test domestic GPU scalability.
Watch for JD's ability to achieve its data collection goals and how it leverages this data to improve physical AI models and applications.
Observe if other companies follow JD's lead in building similar domestic AI infrastructure, which could reshape the competitive landscape in China's AI industry.
