Building a 100,000-Card Domestic Computing Cluster and Launching the JoyAI World Model: JD.com Releases Latest Achievements in Physical AI Development

Building a 100,000-Card Domestic Computing Cluster and Launching the JoyAI World Model: JD.com Releases Latest Achievements in Physical AI Development

On September 9, the JDDiscovery-2026 JD Global Technology Explorers Conference, themed "JoyAI · Leaping into the Physical World," was held in Beijing.

Building a domestic 100,000-card computing cluster, advancing 10 million hours of embodied data collection, launching the JoyAI series world model, introducing the AI shopping assistant "Dongdong," and unveiling the latest lineup of the "Super Brain + Wolf Pack" robot army...

On September 9, the JDDiscovery-2026 JD Global Technology Explorers Conference, themed "JoyAI · Leaping into the Physical World," was held in Beijing. At the conference, JD.com comprehensively showcased its strategic vision and latest achievements in building the "world's largest physical world operations center," launched the "Physical AI Acceleration Plan," and aims to achieve six "global firsts" in the field of embodied intelligence.

JD.com is deeply integrating computing power, data, and model capabilities with its supply chain, applying them across massive scenarios including retail, logistics, technology, healthcare, industry, property development, finance, food delivery, and domestic services. By opening these capabilities to industry and society, JD.com is driving AI from the digital world into the physical world, creating significant value across diverse industries and millions of households.

Building a Super AI Supply Chain to Accelerate the Construction of the "World's Largest Physical World Operations Center"

"JD's AI was not born in academic papers; it was honed order by order on production lines, in warehouses, at delivery stations, and across the entire supply chain process. JD's supply chain is the best training ground and first stop for AI to enter the physical world." At the JDD conference, Cao Peng, Chairman of the Technology Committee and President of JD Cloud at JD.com, articulated the unique value of JD's AI in this way.

Based on its long-accumulated physical scenarios and supply chain capabilities, JD.com proposed building the "world's largest physical world operations center" and creating a "super AI supply chain" to drive AI deep into real-world processes such as production, circulation, and service, enabling it to participate in actual tasks within the physical world.

JD's super AI supply chain is built on six modules: "cloud, data, model, terminal, scenario, and chain," deeply integrating computing power, data, and models with intelligent terminals and real-world scenarios. Scenarios generate data, data trains models, models drive terminals, and task feedback drives continuous AI evolution, forming an "intelligence flywheel." Simultaneously, the supply chain connects R&D, manufacturing, sales, delivery, and maintenance, lowering the barrier for new technologies from validation to productization and large-scale application, forming an "industry flywheel."

Launching the JoyAI-Echo WM World Model with World-Leading Technology

At the JDD conference, JD.com announced the latest progress in three major AI infrastructure areas: computing power, data, and models.

In terms of computing power, addressing physical AI's requirements for real-time response, low latency, and high reliability, JD Cloud has partnered with Moore Threads and others to build a domestic 10,000-card cluster and is planning to construct a 100,000-card cluster. This will serve as a domestically controllable, ultra-large-scale computing foundation in China's AI cloud sector, providing stable computing support for large-scale model training and real-world scenario applications.

In terms of data, JD Cloud's 10 million-hour human data collection initiative—the largest of its kind—is progressing smoothly. JD Cloud has built an embodied data infrastructure covering the entire process of "collection, storage, annotation, training, evaluation, simulation, and testing," transforming human operations in real-world scenarios into urgently needed data "fuel" for embodied model learning and training. Currently, the first open-source dataset, EgoLive, has been made publicly available, with applications from over 8 countries and more than 100 universities and research institutions.

In terms of models, JD.com has formed the JoyAI foundation model matrix covering multimodal models, world models, and embodied models. At JDD, JD.com released JoyAI-Echo WM, a real-time interactive world model. It ranked first with a score of 81.6 in the authoritative public evaluation WBench Navigation for interactive world models, reaching the industry's top level.

Additionally, vertical domain large models such as JD Logistics "Super Brain" 3.0, JD Industry's JoyIndustrial 2.0, and JD Health's Jingyi Qianxun 3.0 were also prominently released. JD Logistics' "Super Brain" large model 3.0 is the logistics industry's first industrial-grade large model application integration that connects the entire chain of warehousing, transportation, and delivery, with unified AI decision-making, coordination, and execution. It takes the lead in achieving the trinity of decision AI, process AI, and physical AI, compressing the solving time for end-to-end optimal route planning of hundreds of millions of parcels from minutes to seconds. The embodied intelligence model achieves a 96.7% execution success rate across multiple logistics task scenarios, ensuring every parcel is delivered to consumers faster and more accurately. JD Industry's JoyIndustrial 2.0 large model can complete mechanical design through simple language commands and rapidly 3D print finished products, significantly improving engineers' design efficiency and user experience.

Accelerating Physical AI Industrial Implementation, Aiming for Six "Global Firsts" in the Robotics Sector

JD's super AI supply chain is not only deeply applied in its own scenarios but also fully open to the public, driving a leap in cost efficiency for AI across diverse industries. Centered on the robotics industry, JD.com launched the "Physical AI Acceleration Plan," empowering the full-link and full-cycle of robot training, R&D, manufacturing, sales, application, and service, aiming to achieve six "global firsts" within a few years:

JD Cloud will build the world's largest embodied intelligence data collection center, collecting over 10 million hours of real-world human scenario video data within two years to solve the "data drought" in embodied intelligence.

Over the next 5 years, JD.com will deploy more than 80 RoboBase robot industrial bases across China, creating the world's largest robot industrial base. This will achieve integrated services for robot display and delivery, maintenance, R&D and design, pilot assembly, data collection, production and manufacturing, and iterative upgrades.

JD Industry initiated the establishment of the "Robot Parts Industry Development Alliance." Over the next 3 years, it will help 100 robot body manufacturers reduce costs across all categories and help 100 component suppliers double their performance, becoming the world's largest robot component supplier.

JD Retail will also invest tens of billions in resources by 2028 to help 100 brands achieve independent sales exceeding 1 billion yuan, driving the robot industry into millions of terminal scenarios and covering tens of millions of users, becoming the world's largest robot retail channel.

Logistics is one of the most promising scenarios for robot applications. JD Logistics is committed to building the world's largest embodied robot application army, purchasing 3 million robots, 1 million unmanned vehicles, and 100,000 drones within 5 years to further advance the unmanned implementation of the entire logistics process. The "Super Brain" large model is responsible for intelligent decision-making and collaborative scheduling in complex scenarios, while the "Wolf Pack" robot army has deployed "nine wolves" in warehousing, delivery, and other processes, including the recently added "Mother Wolf," "Warehouse Wolf," the low-temperature version of "Smart Wolf," and the newly upgraded embodied dexterous arm "Alien Wolf."

JD Logistics will also build the world's largest robot repair service network. Currently, JD Logistics has established 8 major repair centers in China; over the next 5 years, robot after-sales service capabilities are planned to cover more than 100 countries globally, creating over 100,000 jobs for robot after-sales service engineers, providing a guarantee for the large-scale deployment and continuous operation of robot products.

JD App Launches AI Shopping Assistant "Dongdong" to Enhance the Experience of Hundreds of Millions of Users

Beyond penetrating industries, JD's AI capabilities are now serving millions of households, enhancing the consumer experience for hundreds of millions of users.

In retail scenarios, JD.com is further integrating AI capabilities into the complete user journey from demand generation, product selection, and purchasing decisions to fulfillment services. The new version of the JD App released at this year's JDD not only achieves AI enhancement of existing shopping modes but also introduces the AI shopping assistant "Dongdong." Integrating AI's precise understanding of user needs, it shifts from simply helping users find a product to understanding what problems users need to solve, mobilizing massive products and services to meet user needs, making shopping decisions more efficient and the shopping experience more worry-free.

In healthcare scenarios, Jingyi Qianxun 3.0 makes medical AI better at understanding conditions, making judgments, and implementing solutions. JD Health released its first fully intelligent doctor agent, embedding expert diagnostic thinking into AI—users input symptoms, medical history, and examination reports, and it can inquire about the condition, read indicators, and provide medical direction like an expert, while also reminding them about follow-up visits, medication, and check-ups. Meanwhile, through "AI + hardware + service," devices such as blood glucose monitors and blood pressure cuffs are integrated into personal health records, linking online consultations, at-home rapid testing, and nurse visits, turning health management from verbal advice into truly accessible services.

In home scenarios, JD's embodied intelligence brand JoyInside implants AI brains into hardware, driving smart devices from single-product responses to whole-home proactive services. It has currently reached technical cooperation with over 200 brands, covering numerous product categories such as home appliances, home goods, toys, and robots.

Some companies develop AI by first having the technology and then finding scenarios—"making the shoe first and then measuring the foot." JD.com develops AI by growing technology from scenarios—"measuring the foot first and then making the shoe." This is an "inside-out" path: first using AI to optimize its own supply chain efficiency to the extreme, and then opening the polished capabilities to industry partners and customers. Guided by the goal of building the "world's largest physical world operations center," JD.com will act as a "connector" and "operator," working with industry partners to drive AI's leap into the physical world and contributing to the vigorous development of the real economy.

This article is provided by JD.com and republished with authorization by Qubit. The views belong to the original author.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.