Wednesday, November 18, 2026
AI Infrastructure · News & Analysis
Home › Compute & Cloud › Report
Compute & Cloud · Report

JD Cloud plans to build a 100,000-GPU intelligent computing cluster based on Moore Threads GPUs.

A 100,000-GPU domestic cluster would be a notable test of Chinese-made accelerators at hyperscale, signaling demand for home-grown compute alternatives.
Trade pressSlicast · November 17, 2026 at 13:02 UTC · China · Source: 第一财经
importance 65

JD Cloud has announced plans to build a 100,000-accelerator, full-function GPU cluster, creating ultra-large-scale domestic AI computing infrastructure. The cluster, announced at the 2026 JD Global Technology Explorers Conference on September 9, 2026 in Beijing, will be built on Moore Threads' full-function GPUs as its computing foundation. It will focus on large-model training, inference, and key areas such as embodied intelligence, and will open its computing capacity to the industry to help accelerate the industrialization and scaling of physical AI. The company says this marks the first time a domestic GPU has entered the core 100,000-card intelligent computing cluster of a leading AI cloud provider, signaling that domestic computing is moving from "usable" to "large-scale commercial deployment."

Recently, the Ministry of Industry and Information Technology issued the "15th Five-Year Plan for the Information and Communications Industry," which calls for the orderly deployment of intelligent computing clusters with 10,000 cards, 100,000 cards, and above, and for greater efforts to adapt to domestic computing chips. Against this policy backdrop, companies are investing in AI infrastructure with unprecedented intensity, driving continued growth in demand for domestic computing power. According to JD Cloud's plans, the cluster's resources will be prioritized for supply chain AI, embodied intelligence, integrated large-model training and inference, and industrial deployment. In this context, domestic vendors with genuine training-grade GPU capabilities are seen as especially important.

Embodied intelligence, a representative form of physical AI, is approaching its own "ChatGPT moment," and computing infrastructure is the decisive factor. Moore Threads is described as one of the few domestic companies with genuine training-grade GPU capabilities. It has achieved commercial deployment of thousand-card and ten-thousand-card clusters within a single network and has reported breakthroughs in core training scenarios including foundation models, embodied "brain" models, and world models.

Under the partnership, the two companies will collaborate across the full stack, from chips and cloud platforms to model training, to support the evolution and iteration of JD's JoyAI model family, accelerate model training, simulation, and high-quality data generation, and build a complete closed loop of "data → training → simulation → deployment."

Cao Peng, Chairman of JD Group's Technology Committee and President of JD Cloud, said: "We are fully advancing the implementation of our physical AI strategy, and high-performance, highly reliable domestic computing power is the core support. We chose to deepen our cooperation with Moore Threads precisely because of its scarce capabilities in domestic training-grade GPUs and its proven experience in delivering large-scale clusters. This cooperation is an important step in JD Cloud's AI strategy and an important starting point for our embrace of the domestic intelligent computing ecosystem. We sincerely invite more domestic ecosystem partners to join us in rooting domestic computing power in industrial scenarios, so that AI can empower thousands of industries and reach thousands of households at scale."

Zhang Jianzhong, Founder, Chairman and CEO of Moore Threads, said: "Full-function GPUs are the preferred choice for building high-quality domestic computing infrastructure. JD Cloud's selection of Moore Threads for its 100,000-card cluster is the strongest validation of our technical strength and engineering delivery capability, and a key milestone in the move of domestic intelligent computing toward large-scale commercial use. We will work together to build a world-class AI factory, so that domestic computing power supports not only intelligence in the digital world but also enters the physical world, driving industrial transformation in real-world scenarios across all sectors."

Starting from this point, the path to large-scale deployment of domestic GPUs in ultra-large industrial scenarios is now clear, providing a replicable computing paradigm for physical AI to penetrate thousands of industries. As physical AI moves deeper into industry, domestic computing infrastructure with full-chain capabilities is reaching a critical juncture, moving from industrial validation toward large-scale expansion.

About JD Cloud

JD Cloud is the core brand through which JD Group provides technology and services to government and enterprise clients. To meet the digital and intelligent needs of the AI era, JD Cloud has built full-stack AI infrastructure spanning AI cloud, large-model services, and Agent applications. Drawing on JD's deep experience in physical operating scenarios such as retail, logistics, healthcare, and industry, JD Cloud connects the full chain of "AI infrastructure, industrial depth, and user experience," allowing AI to be refined across thousands of real business scenarios. JD is using AI to build the world's largest physical-world operations center, and JD Cloud is opening up this industry-validated capability to help thousands of industries enter a new era of the intelligent economy.

About Moore Threads

Centered on full-function GPUs, Moore Threads is dedicated to providing global accelerated computing infrastructure and one-stop solutions, delivering powerful AI computing support for digital and intelligent transformation across industries. Our goal is to become a GPU leader with international competitiveness and to build advanced accelerated computing platforms for a digital-intelligent world that integrates artificial intelligence and digital twins. Our vision is to accelerate a better world.

Read the original
JD Cloud plans to build a 100,000-GPU… · Slicast