Group 1 - The emergence of "space computing" is driven by the increasing power demand in data centers due to global AI model advancements, leading to the deployment of high-performance satellites in low/mid orbits [1][2] - Space computing offers significant advantages over traditional ground data centers, including higher deployment efficiency, better energy efficiency, and lower cooling costs [1][2] - The "Zhijiang Laboratory + Guoxing Aerospace" collaboration has launched the "Trinity Computing Constellation" with 12 satellites in orbit, aiming for a future capacity of 1000 POPS [1][2] Group 2 - The energy system's weight significantly impacts the overall cost of satellites, with energy costs accounting for 22% of the total satellite economics [2] - Rollable solar arrays combined with flexible batteries are key to the development of space computing systems, with silicon-based HJT batteries being the most suitable for the new generation of rollable solar systems [2] - Companies like NexWafe and Solestial are accelerating their layouts in this area, and HJT batteries are also optimal for perovskite tandem applications, showing long-term evolution potential [2] Group 3 - The current mainstream orbits are LEO and SSO, with SSO providing stable sunlight year-round, making it the best choice for high-power data centers [3] - To address the shortage of orbital resources, space computing platforms are evolving towards large motherships and multi-satellite clusters, with Starcloud constructing a 4km x 4km solar mothership platform [3] - A 10 GW solar capacity can correspond to 448 Google Suncatcher satellites or 2 Starcloud motherships, indicating the scale of deployment needed [3] Group 4 - Investment recommendations focus on companies with overseas customer bases, such as HJT equipment leader Maiwei Co., Ltd. and Gaomei Co., Ltd., which has achieved mass production of 60μm ultra-thin silicon wafers [3]

Investment Rating - The report recommends a focus on HJT equipment leaders with overseas customer bases, specifically highlighting Maiwei Co., Ltd. and Gaomei Co., Ltd. for their production capabilities of ultra-thin silicon wafers [4]. Core Insights - The space computing paradigm is emerging due to the imbalance in supply and demand for computing power driven by AI advancements, leading to the development of "orbital data centers" [3]. - The energy system's weight significantly impacts the overall cost of satellites, with rollable photovoltaic arrays paired with flexible batteries being crucial for development [4]. - The current mainstream orbital resources are limited, prompting the evolution of computing platforms towards large motherships and multi-satellite clusters [5]. Summary by Sections Space Computing - Space computing represents a new paradigm where modular server nodes with training and inference capabilities are deployed on low or medium orbit satellites, creating "orbital data centers" [9]. - The first batch of AI satellites in the "Trinity Computing Constellation" has been launched, with plans for a total of 1,000 satellites [9]. Energy Systems - The cost of energy systems accounts for 22% of the total satellite cost, making it a critical factor in satellite economics [4]. - Rollable structures are gradually replacing traditional Z-shaped structures, becoming the mainstream solution for LEO orbits, and are compatible with flexible, thin-film batteries [4]. Market Dynamics - The limited availability of orbital resources is driving the development of computing platforms towards larger motherships and multi-satellite clusters, with significant deployment plans outlined by companies like Starcloud and Google [5]. - The report emphasizes the importance of HJT technology as the optimal solution for large-scale applications in space computing, particularly due to its compatibility with new energy systems [7]. Investment Recommendations - The report specifically recommends investing in HJT equipment leaders and companies capable of producing ultra-thin silicon wafers, indicating a strong potential for growth in this sector [4].