Chinese e-commerce giant JD.com plans to acquire 3 million robots, 1 million autonomous vehicles, and 100,000 drones over the next five years to automate its logistics network, according to SCMP.
The company also intends to retrain employees to work with robots.
Robots to Replace Couriers and Warehouse Workers
JD.com’s logistics division announced these plans at an event in Beijing, introducing the Wolf Robot industrial line designed for repetitive and labor-intensive tasks.
The robots are expected to handle picking, sorting, and transporting goods, as well as delivering orders with minimal human involvement.
The lineup includes specialized machines for operating in temperatures as low as -20°C, automated systems for delivering medications, and drones.
Currently, JD.com and its affiliates employ approximately 700,000 people in delivery and logistics. As the company scales up automation, it promises not just to reduce human involvement in operations but also to transition some employees to technical roles.
Couriers to Be Retrained to Service Robots
JD.com founder Liu Qiangdong previously introduced the Nirvana Plan, which aims to retrain couriers and warehouse workers to operate robotic systems.
Employees will be able to engage in servicing and repairing robots and perform other technical tasks as machines take over warehouse and delivery operations.
This approach aims to preserve some existing jobs in a new form.
JD.com to Gather 10 Million Hours of Data
In addition to scaling equipment, JD.com is focusing on data to train embodied AI systems that interact with the physical world.
According to JD Logistics Chairman Cao Peng, the company plans to collect over 10 million hours of data on the real-world operations of robots and humans.
JD intends to use its infrastructure, including warehouses, stores, and pharmacies, to achieve this. The company also plans to establish more than 80 RoboBase centers in China for robotics research, pilot production, data annotation, equipment upgrades, and maintenance.
According to SWS Research, the available global volume of quality data on robot-environment interaction was about 500,000 hours at the beginning of 2026. Analysts estimate that tens of millions of hours are needed to train universal embodied AI models.
Computational Cluster for Robots to Be Built
JD.com has also signed an agreement with Chinese chip developer Moore Threads. The companies plan to create a computational cluster based on 100,000 domestic chips.
This cluster will be used for training and deploying AI models and embodied AI applications. JD aims to integrate Chinese processors into the full development cycle—from data collection and model training to simulation and deployment.
In September, Arm introduced the Robotics Capability Framework, a system for a unified description of robot capabilities. Over 80 companies, including AWS, Hugging Face, NXP, QNX, Siemens, Unitree Robotics, Qwen, and others, have already joined the initiative.
