A technological revolution is unfolding in Earth’s orbit as Chinese companies lead the charge to deploy advanced artificial intelligence systems in space. This emerging frontier represents a strategic response to the growing computational constraints facing terrestrial AI development, including energy limitations, physical space shortages, and cooling challenges.
The recent deployment of Starcloud-1 satellite equipped with Nvidia GPUs via SpaceX rocket in early November has brought space-based computing into international focus. Chinese enterprises have positioned themselves at the forefront of this movement, recognizing orbital platforms as a solution to Earth’s AI infrastructure bottlenecks.
Among the pioneers is Zhongke Tiansuan (Comospace), established in 2024, which has achieved a significant milestone with its Aurora 1000 space computer logging over 1,000 operational days aboard a Jilin-1 satellite. The company is preparing to launch its next-generation Aurora 5000 system, featuring domestically developed high-performance GPUs, for orbital trials next year as part of an ambitious project to construct a ‘space supercomputer’ in low Earth orbit.
According to Liu Yaoqi, CEO of Zhongke Tiansuan, orbital edge computing offers distinct advantages by positioning AI capabilities directly at the data source. ‘This approach enables processing petabytes of daily satellite imagery and traffic before transmission through constrained downlink channels,’ Liu explained to Xinhua. Additional benefits include global coverage through low-orbit constellations and nearly free computational power from abundant solar energy.
China’s space computing initiative aligns with broader national ambitions. Beijing municipal authorities recently unveiled plans for a massive orbital data center positioned 700-800 kilometers above Earth in a dawn-dusk orbit. This project, spearheaded by an innovation consortium, targets a system with power capacity exceeding one gigawatt. The initial technology demonstration satellite, Chenguang-1, is scheduled for launch in late 2025 or early 2026, with computing power comparable to a single ground server.
Zhang Shancong, president of Beijing Astro-future Institute of Space Technology (BAIST), which leads the project, acknowledged the modest beginning while emphasizing its significance: ‘Its scale is modest, but we are taking this first small step.’ The deployment strategy involves three progressive phases, culminating by 2035 in a megawatt-scale orbital data center expected to surpass China’s entire current ground-based computing capacity.
In parallel developments, Hangzhou-based Zhejiang Laboratory has established a 12-satellite mini computing constellation named ‘Three-Body,’ equipped with an 8-billion-parameter space-borne AI model. Two satellites within this network carry X-ray polarimeters that combine their computational resources to detect transient gamma-ray bursts in real time. The laboratory projects that upon completion of its planned 1,000-plus satellite constellation, the system will process 100 quintillion operations per second.
‘With a computing constellation, part of the data can be processed in space and delivered straight to users,’ stated Li Chao from Zhejiang Laboratory.
Critical to connecting these distributed orbital computers, China is advancing inter-satellite laser communication technology. Beijing startup Laser Starcom has achieved a breakthrough with a 400 Gbps laser link between its Guangchuan-01/02 satellites, launched aboard a Zhuque-2 rocket last November. Company founder Wu Shaojun emphasized that ‘Laser links are the bedrock that breaks the communication bottleneck and lets space-based computing fly.’
Significant technical challenges remain, particularly regarding operation in extreme radiation environments and heat dissipation in the vacuum of space where conventional cooling methods are ineffective. The Tiansuan team has implemented redundant designs, error correction protocols, and recovery systems to address radiation-induced computational errors and system crashes in industrial-grade chips. They are also experimenting with fluid-loop cooling technology to manage thermal output from high heat-flux components in orbit.
Liu Yaoqi outlined a developmental roadmap beginning with intelligent remote sensing to overcome data transmission limitations, progressing to enhanced communications through large satellite networks for increased capacity and reduced latency, and ultimately evolving toward sophisticated in-orbit AI providing real-time computational support for terrestrial applications. These could include perception systems for autonomous vehicles, drone traffic management, cross-border logistics coordination, and maritime navigation assistance.
Envisioning practical applications, Liu suggested future fisheries might utilize a ‘Fish Finder’ application integrating real-time satellite imagery, environmental data, AIS signals, and on-orbit AI processing to direct fishing vessels precisely to optimal fishing locations.