Chinese scientists and engineers are constructing a space-based computing network to achieve instant processing of satellite-obtained data and build supercomputing capabilities in Earth orbit.
The ambitious initiative, led by the Aerospace Information Research Institute of the Chinese Academy of Sciences, took a concrete step earlier this week with the founding of Xingyun Space AI Technologies, a joint venture by enterprises invested by the academy and other State-owned investors, in Hefei, Anhui province, intending to integrate orbital computing resources with ground infrastructure.
Wu Yirong, an academician of the Chinese Academy of Sciences and academic dean at the Aerospace Information Research Institute, said at an industry forum on Monday that the traditional satellite operating model — capturing imagery or data in space, sending raw files back to Earth, and processing them on ground computers — is characterized by severe time lags and low timeliness.
He cited examples that data collected over oceans or concerning general-aviation and commercial drone operations often cannot reach ground stations in a timely manner, causing delay to data analyses or decision-making.
"Space-based computing solves this by equipping satellites with integrated sensing, computing and communication capabilities, so data can be processed instantly as it is collected. Linked via high-speed inter-satellite connections, these orbital computing assets can work as a complementary network to ground-based computing systems," Wu said.
Speaking at the forum, Shao Zongyou, chairman of Xingyun Space AI Technologies, stressed that space-based computing is not simply moving ground computing hardware onto satellites.
"Instead, it is a complex project that covers remote sensing, semiconductors, space networks, spacecraft thermal management, energy supply and constellation control. Whoever builds the first 'shareable, maneuverable and operable' computing network in space will gain the lead in the space-based information industry," he said.
The executive explained that the "space cloud" model will integrate remote sensing satellites, communication satellites and dedicated computing satellites into one intelligent space-based information service system, covering the entire industrial chain from domestic computing chips to end-user data applications.
According to Shao, Xingyun Space AI Technologies will gradually launch three categories of satellites: remote-sensing satellites with high-performance edge computing capacity, general-purpose computing satellites with strong processing capacity, and supercomputing satellites with ultrafast computing power.
Once completed, he said, the network will enable seamless integration and unified scheduling of space and ground computing resources, allowing users to access space computing services on demand.
When fully operational, the in-orbit infrastructure will form a space information system combining real-time sensing, high-throughput communication and high-performance supercomputing functions.
"It will support public-service sectors including natural resource surveys, emergency and disaster response, as well as low-altitude air businesses, and will be connected to the national computing internet system to provide integrated space-ground computing services to users around the world," Shao said.
Huang Cong, a senior researcher at the National Satellite Meteorological Center in Beijing, said it is very meaningful to equip satellites with on-board computing and processing capabilities.
"Looking ahead, we will deploy ultra-large-scale networks comprising thousands of spacecraft. Within such constellations, satellites are placed relatively close to one another, and they must hold their orbital positions to maintain the constellation's full formation. When space weather conditions change, transmitting adjustment commands via ground uplink would be far too slow to respond in time. Therefore, if the network has built-in on-board computing capabilities — meaning it can process data in orbit and conduct self-adjustment autonomously — it can automatically adjust the satellites' attitude and position to mitigate potential risks," he explained.
Another scenario will be that, when disasters strike, computing-capable satellites can process data in real time in orbit, and then can quickly generate on-site images and analytical results, securing critical time for search and rescue operations, according to Huang.
