Open Source Securities: Aerospace Computing Power Welcomes a Good Opportunity for Layout, Bright Prospects for Space AI

According to the Zhitong Finance APP, Kaiyuan Securities has released a research report stating that bringing computing power to space can achieve on-orbit data processing, improving the timeliness of space data processing, and the prospects for space AI are bright. It is optimistic that the Trisolaris Computing Constellation will solve the bottleneck of traditional satellite data processing efficiency through real-time on-orbit data processing, promoting the application and development of artificial intelligence in space. Recommended investment opportunities in aerospace computing power include CPST (002544.SZ), Geovis (688568.SH), and Piesat (688066.SH).

The main points of Kaiyuan Securities are as follows:

Event: Wang Jian, Director of Zhijiang Laboratory, reiterates the potential of aerospace computing power

On September 11, at the 2025 Inclusion Bund Conference, Wang Jian, founder of Alibaba Cloud and director of Zhijiang Laboratory, delivered a keynote speech titled "Artificial Intelligence: From Code Openness to Resource Openness." He stated that artificial intelligence should not be absent from space. Previously, on May 14, the "Trisolaris Computing Constellation" initiated by Zhijiang Laboratory launched 12 satellites into space and placed the first true 8B AI model on the ground into space, achieving complete interconnectivity of space satellites for the first time.

Bringing computing power to space has great potential, and the prospects for space AI are bright

Under the traditional model, satellites must first transmit data back to the ground, where ground data processing centers analyze it. This "sky sensing and ground computing" model is limited by ground station resources, bandwidth, and other factors, resulting in less than one-tenth of effective satellite data being transmitted back to the ground, along with issues such as poor data timeliness. According to Li Chao, chief engineer of the overall scientific research task of the computing constellation and deputy director of the Space-Based Computing System Research Center, bringing computing power to space can achieve on-orbit data processing, allowing a large amount of data to be processed in real-time without needing to be transmitted down, thus improving the timeliness of space data processing and making it possible to move intelligence to space. It is optimistic that the Trisolaris Computing Constellation will solve the bottleneck of traditional satellite data processing efficiency through real-time on-orbit data processing, promoting the application and development of artificial intelligence in space.

Industry trends are accelerating, and future potential is worth looking forward to

The 12 computing satellites launched into orbit in May are equipped with onboard intelligent computing systems and inter-satellite communication systems, enabling interconnectivity of the entire satellite constellation and providing on-orbit computing capabilities. The maximum single-satellite computing power of the computing satellites reaches 744 TOPS (trillions of operations per second), and the maximum inter-satellite laser communication rate can reach 100 Gbps (bits per second). After interconnecting the 12 satellites, they will have a computing capacity of 5 POPS (quintillion operations per second) and a storage capacity of 30 TB. The satellites are also equipped with a space-based model with 8 billion parameters, capable of on-orbit processing of L0-L4 level satellite data, and will perform on-orbit experimental tasks such as laser access for non-cooperative satellites and astronomical observations. According to the plan, in 2025, Zhijiang Laboratory will collaborate with Guoxing Aerospace, Zhixing Space, and other companies to jointly complete the constellation layout of over 50 computing satellites. Ultimately, the goal of the "Trisolaris Computing Constellation" is to create a space-based intelligent computing infrastructure on a scale of thousands of stars, with a total computing power scale expected to reach 1 EOPS (hundreds of billions of operations per second), which is anticipated to exceed most single data centers currently on the ground Risk Warning: The implementation of the industry is not as expected; the risk of intensified market competition; the risk of the company's research and development not meeting expectations, etc