Industry Insights 12 min read

JD.com Unveils 100K-Card Domestic AI Cluster, JoyAI World Model & 6 Robotics Global Firsts

At its 2026 global tech conference, JD.com revealed a 100,000-card domestic compute cluster, the JoyAI-Echo world model scoring 81.6 on WBench Navigation, 10 million hours of embodied data collection, and a Physical AI Acceleration Plan targeting six global firsts in robotics data, manufacturing, parts supply, retail, deployment, and maintenance.

JD Tech Talk
JD Tech Talk
JD Tech Talk
JD.com Unveils 100K-Card Domestic AI Cluster, JoyAI World Model & 6 Robotics Global Firsts

At the JDDiscovery-2026 conference on September 9, 2026, themed "JoyAI · Leap into the Physical World," JD.com showcased its latest progress toward building the "world's largest physical world operation center" and launched the "Physical AI Acceleration Plan" targeting six "global firsts" in embodied intelligence.

Super AI Supply Chain: Six-Module Foundation and Dual Flywheels

JD.com's technology committee chair and cloud president Cao Peng emphasized that JD's AI originates from production lines, warehouses, and delivery stations — "measure feet before making shoes" rather than "make shoes then measure feet." The super AI supply chain rests on six modules: cloud, data, model, terminal, scenario, and chain. Scenarios generate data, data trains models, models drive terminals, and task feedback fuels continuous AI evolution, forming an "intelligent flywheel." Simultaneously, the supply chain connects R&D, manufacturing, sales, delivery, and maintenance, lowering barriers from verification to productization to scale, creating an "industrial flywheel."

Diagram of JD's super AI supply chain six modules and dual flywheels
Diagram of JD's super AI supply chain six modules and dual flywheels

AI Infrastructure Advances: Compute, Data, and Models

Compute: Domestic 100K-Card Cluster

To meet physical AI demands for real-time response, low latency, and high reliability, JD Cloud has built domestic 10K-card clusters with partners like Moore Threads and plans a 100K-card cluster, establishing an autonomous, ultra-large-scale compute foundation for massive model training and real-world deployment.

Data: 10 Million Hours of Embodied Data Collection

JD Cloud is advancing a 10-million-hour human embodied data collection initiative, the largest of its kind. It has built full-pipeline embodied data infrastructure covering collection, storage, labeling, training, evaluation, simulation, and testing, converting real-world human operations into training "fuel" for embodied models. The first open-source dataset, EgoLive, has been released and adopted by over 100 universities and research institutions across eight countries.

Models: JoyAI Matrix and JoyAI-Echo WM Benchmark Leadership

JD has formed a JoyAI foundation model matrix spanning multimodal, world, and embodied models. The newly released real-time interactive world model JoyAI-Echo WM achieved 81.6 points on the authoritative WBench Navigation benchmark for interactive world models, ranking first globally.

JoyAI-Echo WM benchmark result on WBench Navigation
JoyAI-Echo WM benchmark result on WBench Navigation

Additionally, vertical domain models were upgraded: JD Logistics "Super Brain" 3.0 unifies warehousing, transportation, and delivery decision-making, reducing billion-package optimal routing solve time from minutes to seconds, with embodied model multi-task success rate reaching 96.7%. JD Industrial JoyIndustrial 2.0 enables mechanical design via natural language commands and rapid 3D prototyping. JD Health "Jingyi Qianxun" 3.0 introduces an expert-level AI agent that mimics specialist diagnostic reasoning and integrates with hardware for closed-loop health management.

Physical AI Acceleration Plan: Six Global Firsts in Robotics

JD opened its super AI supply chain to the industry, launching the Physical AI Acceleration Plan to empower the full robot lifecycle — training, R&D, manufacturing, sales, application, and service — with six targets to achieve within years:

World's largest embodied intelligence data collection center: Collect over 10 million hours of real-scene video data within two years, solving the embodied data scarcity.

World's largest robot industrial base: Deploy 80+ RoboBase bases nationwide in five years, integrating showcase, delivery, maintenance, R&D, pilot assembly, data collection, manufacturing, and iterative upgrades.

World's largest robot parts supplier: Form a robot parts industry alliance to help 100 robot OEMs cut costs across all categories and double revenue for 100 component suppliers within three years.

World's largest robot retail channel: Invest 10 billion RMB by 2028 to help 100 brands exceed 1 billion RMB independent sales, pushing robots into one million terminal scenarios covering ten million users.

World's largest embodied robot application army: Procure 3 million robots, 1 million unmanned vehicles, and 100,000 drones in five years, advancing full-process logistics automation. The "Super Brain" model handles complex decision-making and scheduling, while the "Wolf Pack" robot fleet now includes nine variants such as "Mother Wolf," "Warehouse Wolf," "Smart Wolf" low-temperature version, and the upgraded dexterous arm "Alien Wolf."

World's largest robot maintenance service network: Expand from eight domestic repair centers to cover 100+ countries in five years, creating over 100,000 robot service engineer jobs to support scaled deployment and continuous operation.

Wolf Pack robot lineup and logistics deployment
Wolf Pack robot lineup and logistics deployment

Consumer-Facing AI: Shopping Assistant "Dongdong," Health, and Smart Home

JD's AI capabilities also serve hundreds of millions of consumers. The new JD App integrates AI across the entire shopping journey and introduces the AI shopping assistant "Dongdong," which moves beyond product search to understand user problems and orchestrate vast goods and services for efficient, worry-free decisions. In health, "Jingyi Qianxun" 3.0 deploys an expert AI agent that replicates specialist reasoning — asking symptoms, reading indicators, suggesting care directions, and reminding follow-ups — while connecting glucose meters, blood pressure monitors, and other devices to personal health records, linking online consultation, at-home testing, and nurse visits. In the home, the JoyInside platform embeds AI brains into hardware, shifting from single-device response to whole-home proactive service, with technical partnerships across 200+ brands covering appliances, furniture, toys, and robots.

JD's approach is "inside-out": first maximize its own supply chain efficiency with AI, then open the refined capabilities to industry partners and customers. Guided by the "world's largest physical world operation center" vision, JD positions itself as a "connector" and "operator" to drive AI's leap into the physical world and fuel real economy growth.
Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Supply ChainRoboticsEmbodied IntelligenceAI InfrastructureWorld ModelJD.comPhysical AIJoyAI
JD Tech Talk
Written by

JD Tech Talk

Official JD Tech public account delivering best practices and technology innovation.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.