The transition from concept to engineering raises the question: when will the space computing industry reach its inflection point? At the Space Computing Industry Conference held on April 3, 2026, the description of an app called "Fish Code" illustrated a vision for widespread access to space-based computing power. Liu Yaoqi, an associate researcher at the Institute of Computing Technology of the Chinese Academy of Sciences, shared a student's idea: a fisherman using the app to ask, "Where are the tuna?" Subsequently, satellites equipped with hyperspectral cameras would locate the fish, an intelligent system would perform calculations, and communication links would provide answers including location and fishing gear recommendations. This seemingly sci-fi scenario is expected to become reality sooner as "space computing" progresses from concept to engineering. At the conference, representatives from government, industry, and academia discussed the challenges and pathways for "sending computing power to space." According to multiple industry insiders, the commercialization of space computing in China currently faces several challenges, including key technologies and economic costs. The industry is seeking breakthroughs through technological innovation and business model transformation.
"Computing, Communication, Thermal Management, and Energy" Present Significant Challenges What is space computing? Industry experts explain that it refers to building space-based information infrastructure by deploying computing systems, data storage systems, and high-speed data interconnection facilities in orbit, integrating computing power, storage capacity, and connectivity. In the traditional model, satellites must transmit data back to Earth for processing by ground data centers, known as "space data, ground computing." Under the space computing framework, satellites become "computers with wings," capable of real-time in-orbit data processing and autonomous decision-making. Li Jie, Deputy Director of the Cloud Computing and Digitalization Research Institute at the China Academy of Information and Communications Technology, outlined three stages of space computing development: space data processed in space, ground data processed in space, and space-based dominant computing. Currently, space computing is in the exploratory "space data, space computing" stage, transitioning from proof-of-concept to early engineering. Space computing gained significant attention starting in the second half of last year. Its popularity is driven by the AI wave's demand for massive data processing, surging energy consumption of ground data centers, breakthroughs in technical verification, and the introduction of supportive policies. However, deploying computing power in space is not easy. Liu Jingjing, Chief Operating Officer of Guoxing Astronautics, stated in an interview that "computing, communication, thermal management, and energy" are key challenges. For computing, high-performance radiation-resistant computing chips and payloads need to be developed. For communication, high-speed, stable laser links between satellites and between satellites and ground must be achieved. For thermal management, technologies for ultra-high heat flux collection and large-area heat dissipation are required. For energy, large-scale new energy power supply systems must be constructed. Liu Yaoqi elaborated on issues like "radiation-resistant computing chips" and "thermal management." For instance, radiation can cause single-event upsets and latch-ups, leading directly to data errors in chips. Additionally, vacuum and extreme temperature differences can cause material fatigue and performance drift.
He mentioned that in the vacuum environment without air convection, conventional air cooling methods are completely ineffective. Today, a high-performance AI chip can consume hundreds of watts, with heat flux density far exceeding that of traditional aerospace-grade chips. This necessitates reliance on more complex liquid circulation cooling systems, which introduces new systemic engineering challenges. "From how to export heat from the chip, to the selection of soft or hard thermal pads, to the microchannel design of cold plates, the long-term stability of coolant, and the reliability of circulation pumps, every step is like walking on thin ice. It is a systemic scientific problem requiring extensive experimental validation," he explained. For example, an in-orbit computer project with computing power of 3 petaflops repeatedly failed during ground testing due to a tiny, almost undetectable air bubble, causing a year-long delay. Liu Yaoqi also believes that the application ecosystem in space is still in its infancy, and building an ecosystem for space information is urgently needed.
Multiple Technical Pathways Under Exploration Facing harsh physical challenges and vast market prospects, global explorers are demonstrating diverse technical approaches spanning system architecture, chips, energy, thermal management, and launch vehicles. Regarding system architecture, Xiang Jiying, Chief Scientist at ZTE Corporation, summarized three main paths. The first is the "on-orbit cluster" approach explored by Google. It involves flying several satellites in formation at extremely close distances of a few hundred meters in a dawn-dusk orbit that avoids Earth's shadow. The proximity allows high-speed laser links to create a network similar to a ground data center's internal network, supporting AI model training and inference in orbit. This scheme requires extremely precise formation control and has a high technical barrier. The second is the "distributed computing" path represented by Musk's Starlink. This approach relies on tens of thousands of communication satellites, each with relatively weak computing power, widely distributed. This architecture suits low-latency inference tasks but struggles to support the massive data exchange and parameter synchronization required for AI training, as bandwidth and latency in distributed systems become bottlenecks. The third is the "space supercomputing center" concept, still on paper in Europe. The idea is similar to building a "computing space station," assembling a large, centralized supercomputer in orbit through multiple launches. Considering national conditions and industrial characteristics, Xiang Jiying suggested following the second path, distributed computing. "The entry barrier is relatively lower, and shortcomings in individual satellite quality can be compensated for by launching a larger number of satellites." Regarding chips, industry insiders proposed approaches like lightly customized commercial chips, specialized radiation-resistant chips, and space-native chips. Xiang Jiying mentioned that both NVIDIA and Google adopt lightly customized ground-based chips, a path also suitable for China. Liu Yaoqi proposed a more advanced idea: leveraging the space environment itself to design new materials and devices. Perhaps future space computers shouldn't just "resist radiation" but "absorb radiation." Regarding thermal management, active thermal control was frequently mentioned at the conference. For instance, Galaxy Space已验证了泵驱动散热系统 on flat-panel satellites launched in 2023. The Institute of Computing Technology is tackling systemic engineering issues like microchannel design and pump design. Building an industrial ecosystem is also on the agenda. The conference witnessed the establishment of the "Space Computing Professional Committee" under the Computing Power Industry Development Alliance. As the first nationwide specialized collaborative platform, the committee brings together academicians, experts, leading enterprises, research institutes, and financial institutions from the industrial chain. Li Jie stated that the committee's establishment will enhance synergy between the computing power and aerospace industries, creating an integrated industrial ecosystem.
The Cost Equation The space computing industry requires support from multiple technological segments, implying that the deployment cost of space computing is high. How does the cost breakdown look? Song Zhengji, a researcher at the Beijing Institute of Spacecraft System Engineering, conducted relevant research. He broke down the cost of putting computing power in space: launch costs account for about 30%-40%, satellite manufacturing costs about 20%-30%, with space environment adaptation and computing chips and energy systems each constituting significant portions. Building a 30-megawatt data center in space would still cost roughly an order of magnitude more than its ground-based counterpart. Under these circumstances, when will the space computing industry reach its inflection point, and how will it achieve a commercial closed loop? Reducing rocket launch costs is a consensus among industry players. Representatives from companies like Landspace and Interstellar Glory mentioned mastering rocket recovery technology to achieve reusability. "If industrialization proceeds smoothly, with a design allowing first-stage rockets to be reused 20 times, launch costs could drop to about 20,000 RMB per kilogram," said Xie Hongjun, Deputy General Manager of Interstellar Glory Group. He also predicted that achieving two-stage reusability could potentially balance ground-based and space-based costs. Beyond launch costs, factors like the mass production of perovskite photovoltaics and decreasing costs of commercial chip hardware are also important drivers for space computing development. The industry's inflection point seems not far away. Some institutions predict the global space computing market will exceed one hundred billion US dollars by 2030. Multiple industry insiders stated that there is an urgent need to develop space computing in areas like national security, the low-altitude economy, ocean monitoring, and information services. The application scenarios likely to first achieve a commercial closed loop are those requiring extreme real-time performance, areas where ground networks have poor coverage or are too costly, such as Earth observation and remote sensing, including emergency response, security, and environmental monitoring. "Compared to ground data centers, the advantages of space computing lie in 'real-time capability' and 'coverage,'" added Xie Lina, Deputy Director of the Data Center Department at the CAICT Cloud Computing and Big Data Research Institute. Computing satellites can achieve seamless global coverage through laser communication networks, process data directly in orbit, and transmit high-value information back, reducing data latency for scenarios like disaster warning and resource monitoring. Tianyi Space, a commercial SAR remote sensing satellite constellation operator, believes computing power will define the next phase of commercial aerospace. The company's co-founder and CTO, Ren Weijia, said they have been continuously enhancing on-board computing power and are collaborating with Beihang University to increase it to about 200 tokens. This will improve remote sensing service response times from days to sub-hours, aiming for minutes in the future, enabling effective disaster warnings. "The stronger the space computing power, the wider the application boundaries will expand. The two are currently forming a positive feedback loop," Ren Weijia believes. "In the next five years, space computing will transition from a 'luxury' to a standard infrastructure for global sensing networks."
Comments