JD Cloud Picks Moore Threads Chips for a 100,000-GPU Supercomputer

JD Cloud, the cloud computing arm of JD.com, announced on September 9, 2026, plans to build a 100,000-GPU cluster powered by Moore Threads chips, marking the first deployment of Chinese GPUs at that scale by a major domestic AI cloud provider. The cluster, intended for large-model training, inference, embodied AI, and robotics workloads, will be opened to companies across industries. Moore Threads, a Beijing-based chipmaker founded by former Nvidia vice-president Zhang Jianzhong, listed on Shanghai's STAR Market in December 2025 and reported 2025 revenue of 1.505 billion yuan, up 243.37% year-over-year, with a net loss of about 1.001 billion yuan.

JD Cloud says it will build a 100,000-GPU computing cluster on Moore Threads chips, but the real test is not the headline number. It is whether China's domestic GPU stack can run real cloud workloads at that scale. JD Cloud, the cloud computing arm of JD.com, announced on September 9, 2026, that it plans to build a 100,000-GPU cluster powered by Moore Threads. That is a big bet on a young supplier. Moore Threads is a Beijing-based chipmaker that listed on Shanghai's STAR Market in December 2025, less than a year before this announcement. JD unveiled the plan at its 2026 Global Technology Explorers Conference, framing it as the first deployment of Chinese GPUs in a 100,000-GPU core cluster by a major domestic AI cloud provider, according to TechNode. The cluster is meant for large-model training, inference, embodied AI and robotics workloads. JD Cloud says the computing capacity will be opened to companies across industries, according to AASTOCKS Financial News, which carried the September 9 announcement. The two companies are not starting from zero. Chinese financial news reports say JD Cloud previously built a 10,000-GPU domestic cluster with partners including Moore Threads, so this is a scale-up of a stack JD has already tested, not a cold bet. This is not a science project. Moore Threads is still new to public markets. Founded by Zhang Jianzhong, a former Nvidia vice-president and head of Nvidia's China operations, the chipmaker priced its Shanghai IPO at 114.28 yuan a share and closed its first day at 600.5 yuan, according to market reports on the listing. That's a strong debut. TrendForce, citing the company's annual report, said Moore Threads generated 1.505 billion yuan in 2025 revenue, up 243.37% from the year before, while still posting a net loss attributable to shareholders of about 1.001 billion yuan. Investors are treating it as one of China's best answers to Nvidia in a market where Nvidia's strongest AI chips are no longer freely available. That's the bet. Instacart Launches Clementine, an AI Assistant That Builds Your Cart for You https://startupfortune.com/instacart-launches-clementine-an-ai-assistant-that-builds-your-cart-for-you/ Instacart launched Clementine on September 9, 2026, an AI shopping assistant that turns a chat, recipe, or photo of a handwritten list into a ready-to-buy cart. Early orders placed through it are already topping the company's $113 average basket size, arriving months after Uber Eats and DoorDash rolled out their own AI shopping assistants. - AI assistant that builds grocery shopping carts automatically https://startupfortune.com/instacart-launches-clementine-an-ai-assistant-that-builds-your-cart-for-you/ - how to use Instacart Clementine AI shopping feature https://startupfortune.com/instacart-launches-clementine-an-ai-assistant-that-builds-your-cart-for-you/ China wants more than one Nvidia substitute JD Cloud is not the first Chinese player to reach for the 100,000-card number. Sugon unveiled its own fully domestic 100,000-card supercluster, the Sugon 8000 Dengfeng, at the World Artificial Intelligence Conference in Shanghai in July 2026. Same round number, different chip. That system runs on Hygon chips rather than Moore Threads GPUs, uses Sugon's scaleFabric interconnect, and is designed to connect with China's national supercomputing infrastructure. Two different chipmakers, two state-linked cloud pushes, one identical number. That is not a coincidence. It is a signal that Beijing wants proof, at scale, that its chip industry can stand in for Nvidia across more than one vendor. Nvidia's position in China has been shrinking for a while. In its fiscal 2026 first-quarter filing, Nvidia said H20 sales reached $4.6 billion before new U.S. export licensing requirements took effect on April 9, 2025. In the next quarter, Nvidia said there were no H20 sales to China-based customers. That is a collapse, not a dip. JD Cloud running flagship AI infrastructure on Moore Threads instead of Nvidia is the kind of decision a major, consumer-facing Chinese company only makes once it thinks domestic chips can actually do useful work. It is not just a political gesture with no operational teeth behind it. The missing proof is performance Nobody outside JD and Moore Threads knows yet whether the 100,000-GPU plan will work as advertised, and neither company has published independent benchmarks comparing this cluster's real-world training throughput with an Nvidia deployment. No completion date. No site list. No power draw, customer launch date or full-cluster reliability data. None of it has been made public. That is the open question that will decide whether this project becomes a genuine substitute or a subsidized showcase. Moore Threads has made narrower claims about its own hardware. Its official product materials say the MTT S5000 is built for AI training, inference and high-performance computing, and that its software stack supports frameworks such as PyTorch, Megatron-LM, vLLM and SGLang. Those claims matter because software is where many chip challengers lose the room. You can buy silicon. You cannot fake a developer ecosystem for long. Frankly, the timeline is the tell here. JD Cloud has moved from a 10,000-GPU domestic cluster to a plan for ten times that scale just as the squeeze on Nvidia's China business has tightened. China's chip decoupling from Nvidia stopped looking like a temporary disruption in 2025, when H20 sales to China-based customers went from billions of dollars to zero in one quarter. It is now how the country's biggest cloud buyers have to plan. The practical test comes next. If JD Cloud can turn the Moore Threads cluster into rentable capacity for model builders, robotics teams and industrial AI users, the announcement will look like an early marker in China's domestic cloud stack. If it cannot publish reliable performance and customer use, the number will stay what it is today: a plan with serious strategic weight, but not yet proof. Anthropic Researcher Jacob Coxon Quits Warning AI Could Kill Us All https://startupfortune.com/anthropic-researcher-jacob-coxon-quits-warning-ai-could-kill-us-all/ Jacob Coxon, a 27-year-old pretraining researcher, resigned from Anthropic on September 8, 2026, warning that AI labs are racing toward self-improving superintelligence they can't control. Anthropic's own Alignment Science Lead, Evan Hubinger, publicly agreed there's more than a 10% chance AI kills all humans within a decade. The warning lands as... - Anthropic researcher quits over AI safety concerns https://startupfortune.com/anthropic-researcher-jacob-coxon-quits-warning-ai-could-kill-us-all/ - AI superintelligence risk researcher leaves tech industry https://startupfortune.com/anthropic-researcher-jacob-coxon-quits-warning-ai-could-kill-us-all/ Also read: Massachusetts Now Makes AI Data Centers Bring Their Own Clean Power https://startupfortune.com/massachusetts-now-makes-ai-data-centers-bring-their-own-clean-power/ • Xiaomi's Xring O3 Chip Tops 5 Million AnTuTu Points Before Apple's Fold Debut https://startupfortune.com/xiaomis-xring-o3-chip-tops-5-million-antutu-points-before-apples-fold-debut/ • OpenAI Cuts ChatGPT Image Generation Time in Half With Images 2.5 https://startupfortune.com/openai-cuts-chatgpt-image-generation-time-in-half-with-images-25/ This article is posted in Technology News https://startupfortune.com/category/technology/ , check it out for more related stories. Join the discussion Open in the community → /community/ Almost there. Sign in and your reply posts straight away.