During the 2026 JD.com 618 kickoff conference on May 18th, Cao Peng, Chairman of the JD.com Technology Committee and President of JD Cloud, announced that JD.com will establish the world's largest embodied artificial intelligence data collection center. The initiative will mobilize 600,000 participants in what is described as the largest-scale data collection effort in human history, aiming to accumulate 10 million hours of real-world scenario video data within two years. This effort is designed to create an ecological closed loop encompassing "data collection, model training, and data optimization."
Cao Peng emphasized that JD.com will continue to increase its investments to solidify its AI infrastructure. In 2026, the company's AI-related research and development expenditure within the JD ecosystem is projected to grow by over 200%. Driven by this unprecedented resource commitment and over two decades of accumulated supply chain capabilities, JD.com aims to build a robust foundation for AI to advance into the physical world.
At the data infrastructure level, the embodied AI industry currently faces a critical bottleneck: a scarcity of real-world scenario data, which leads to insufficient model training. While robots may possess advanced "cerebellar" functions for physical actions like dancing or boxing, the "cerebral" embodied models lag behind, hindering deeper industrial application of the technology. Leveraging its more than 20 years of supply chain expertise and diverse business scenarios, JD.com possesses vast amounts of authentic data from areas such as product transactions, logistics fulfillment, and customer service. This data serves as "high-quality fuel" for JD.com's AI research, development, and application.
Capitalizing on these strengths, JD.com is addressing industry pain points by launching the world's first embodied AI data infrastructure that covers the entire chain of "collection, storage, labeling, training, evaluation, simulation, and testing." This infrastructure is intended to provide embodied models with massive volumes of high-quality, real-world data, helping the models learn to "understand" the physical environment.
Cao Peng stated that JD.com is starting by building this data infrastructure to propel the evolution of the industry's embodied AI models. The company is also opening up its JoyInside embodied intelligent interaction capabilities to equip robots with dedicated "intelligent brains." By 2026, JoyInside is expected to be integrated into over 10 million smart hardware devices. During the upcoming JD.com 618 shopping festival, a suite of AI-powered home scenario products featuring JoyInside will debut. For instance, the Xinfei AI Voice Air Conditioner will have its exclusive launch on JD.com, significantly lowering the usage barrier for elderly and child users. Products like AI mattresses, AI wheelchairs, and AI desk lamps will also greatly enhance user convenience.
In service consumption sectors such as healthcare, food delivery, and home services, clusters like the AI Jingyi Intelligent Agents and the food delivery AI network-wide review large model will comprehensively serve over 700 million users. All these products are set to offer users novel experiences during the JD.com 618 period.
Furthermore, JD.com is accelerating the commercialization of the robotics industry, including sales, leasing, and after-sales maintenance, while providing robotics companies with comprehensive industrial and production development supply chain support services. This year, JD.com aims to help robotics brands achieve a cumulative sales scale exceeding 10 billion yuan and shorten product time-to-market by 30%.
Through these measures, JD.com has established a full-chain embodied AI ecosystem within the broader industry. In the future, when users place orders on JD.com, robots will provide end-to-end services, from product procurement and warehousing to handling, delivery, and in-home fulfillment. Such scenarios will serve as a vivid illustration of JD.com's ambition to build the world's largest physical world operation center.