Shanghai Xingshu Tiansuan Space Technology Co. announced on Saturday that it has launched the first constellation of its Tiansuan Constellation project, a phased plan that aims to place 1,000 satellites in orbit to process artificial intelligence and remote-sensing workloads directly in space. Reuters reported the announcement was made at a computing power forum held during the 2026 World AI Conference in Shanghai.
The project, known formally as the Xingshu Plan, takes its name from the Chinese phrase meaning 'central hub of the stars.' It is designed to deploy two categories of satellite: full computing satellites and smaller edge-computing satellites, which handle localised data processing tasks. According to China.org.cn, the verification phase involves launching two computing satellites and 12 edge-computing satellites, before expanding to 50 and 100 respectively in the commercial phase, and eventually reaching a network of approximately 1,000 satellites.
Why process data in orbit at all?
Space-based computing refers to running AI models and data analysis tasks aboard satellites rather than relaying raw sensor data back to Earth. The logic is straightforward: Earth-observation satellites and other sensing platforms generate far more information than they can practically transmit to ground stations, particularly during the narrow windows when a satellite is within range. Researchers at Zhejiang Lab have estimated that up to 90 percent of data generated in space is currently lost and never effectively processed, largely because of bandwidth constraints.
There are also incentives on the ground side. Global data centres already consume around 1.5 percent of the world's total electricity generation, with roughly 40 percent of that energy going purely to cooling systems. Processing certain workloads in orbit, where solar power is consistent and heat can be radiated into the vacuum of space, sidesteps some of those physical limits. Additionally, the vast majority of the Earth's surface, including 70 percent of its land and 95 percent of its oceans, falls outside existing network coverage, making in-orbit computing a potential solution for remote operations.
A crowded field, with geopolitical weight
Xingshu Tiansuan is not operating in isolation. China has several companies pursuing comparable orbital compute projects, including Zhongke Tiansuan and Zhejiang Lab's Three-Body Computing Constellation, which placed its first 12 satellites in orbit in May 2025 and targets a 100-satellite network by 2027. Beijing has also moved to coordinate the sector institutionally, approving a Space Computing Industry Innovation Center that draws together satellite manufacturers, semiconductor firms, AI companies, and universities under a single policy framework, aligned with priorities in China's 15th Five-Year Plan running to 2030.
“In the future, space-based computing power will become a universal public service, just like water and electricity.”
The timing of Saturday's announcement was not incidental. President Xi Jinping was attending the 2026 World Artificial Intelligence Conference in Shanghai, running from 17 to 20 July, where he positioned Beijing as a leading voice shaping global AI governance. The Xingshu launch formed part of a broader showcase of Chinese AI ambition during the conference week.
Internationally, the competition is intensifying. SpaceX, which completed a full merger with Elon Musk's AI company xAI in February 2026, is advancing its own space-based computing initiatives, with Musk having described orbital data centres as a solution to the energy and cooling limitations constraining AI growth on Earth. Amazon's Jeff Bezos is also pursuing the 51,600-satellite Project Sunrise. What distinguishes China's approach, analysts note, is its coordinated industrial model: rather than a single vertically integrated company, Beijing is assembling an ecosystem of specialised firms working toward a shared infrastructure. Whether that model proves more durable than the concentrated bets being placed in the United States remains one of the more consequential questions in the emerging space-computing race.
This article is free to read. It always will be — no paywall, no account, no tracking.
