Robot Forward reported on 9 September that at the JDD 2026 JD Global Technology Explorer Conference, JD pointed the next decade of AI at the physical world. Cao Peng, chairman of JD’s technology committee and president of JD Cloud, said the smart economy will migrate from the digital world to the physical world over the coming ten years.
Where digital and smart-economy value was built around recommendation, dialogue, content, marketing and office work, AI entering the physical world changes more than single-point efficiency: it reshapes how goods are made, moved and consumed. JD’s answer is a Super AI Supply Chain that links data, model, cloud, device, chain and field, pushing AI from the screen into the real world, backed by six Wolf embodied products.
From single-point rivalry to a system play
Cao framed physical AI as moving from a contest of models and bodies toward systematic competition. On data, JD is scaling collection and will open a vast first-person human-operation video dataset called EgoLive, targeting more than 10 million hours of collected and labelled data by the end of 2027. This year it has already open-sourced close to 2,000 hours across 346 real tasks.
On models, JD built the JoyAI matrix stretching from general language, code, multimodal and audio-visual ability toward vertical and embodied intelligence. Its real-time interactive world model JoyAI-EchoWM scored 81.6 to rank first on the public WBench Navigation benchmark. For embodied work it fields navigation model JoyAI-Nav, manipulation model JoyAI-RA, spatial reconstruction model JoyAI-Spatial and physics simulator JoyAI-Sim.
On compute, JD Cloud and Moore Threads are building a domestic ten-thousand-card cluster, with a planned hundred-thousand-card AI factory to underwrite large-scale training. On device, field and chain, JD will lay out more than 80 robot industry bases nationally, JD Logistics will buy 3 million robots, 1 million unmanned vehicles and 10,000 drones over five years, and JD Retail will put 10 billion yuan of resources behind the push by 2028.
Six Wolves take real jobs
The six Wolf products are Zhi Lang, Fei Lang, Yi Lang, Du Lang, Cang Lang and Mu Lang, covering storage, picking, sorting, transport and delivery. Yi Lang targets the hardest parcel sorting, judges how to grasp and place items of varying shape, material and weight by itself, and has already completed more than 13 million real parcel operations on 24-hour unmanned runs.
Cang Lang retrofit existing warehouses by fitting current racks and aisles instead of rebuilding them. Zhi Lang’s low-temperature version takes robots into cold-chain work, fighting material brittleness, drivetrain slip and frost. Mu Lang health edition runs pharmacy storage, management and dispensing as an automated flow with full traceability on 24-hour runs.
Du Lang carries high-frequency transport between warehouse and station using L4 driverless capability, pushing autonomy from inside the building out to public roads. Fei Lang extends unmanned delivery into low-altitude air. JD also showed an AI smart park spanning more than 28 million square metres that unifies robot dogs, cleaning robots and delivery robots onto shared maps and joint tasks.
More images from the source report:
Editor’s note: This is an adapted translation of the original Zhidx report. It has been trimmed and restructured for readability for an international business audience.
Translated and adapted from Zhidx (https://www.zhidx.com/p/592211.html).