Guosheng Securities: Space computing power has gradually moved from conceptual verification to engineering implementation stage.

Guosheng Securities released a research report stating that with the announcement of large projects such as "Stargate" and the continuous upward revision of CSP manufacturers' Capex, the explosive growth in computing power demand has become a certainty trend, but ground infrastructure may face bottlenecks such as energy, cooling, and land. The bank pointed out that space computing power has become a strategic solution to address the challenges of energy and cooling faced by ground computing power. Companies like Nvidia, Amazon, and other AI giants have already started positioning themselves in this trend, and the participation of domestic and foreign giants and policy formulation may accelerate the development of this field, and related investments will begin to emerge. Guosheng Securities' main points are as follows: At the current point in time, we continue to look for other extensions of AI infrastructure. The market is focused on hardware facilities such as optical modules and switches in the computing power infrastructure, or expecting breakthroughs in AI applications. However, when computing power scales surpass a critical point, the underlying logic of energy, space, and cooling may have undergone a qualitative change. As the power demand of data centers surges and the US power infrastructure lags behind, the real-time response of the US Starcloud gigawatt-level orbital power plant and China's "Three-Body Computing Constellation" is indicating that the computing power battlefield may have extended from the ground to the sky. What is space computing power? - Data centers deployed in space orbits. Space computing power is a technology that deploys data centers and computing capabilities to space orbits, processes data on orbit using satellites and their computing hardware, utilizes inter-satellite high-speed laser communication for data transmission and real-time processing, and transmits results back to Earth. Due to the unique vacuum environment and lighting conditions in space, space computing power features autonomous intelligence, real-time response, distributed cooperation, and no energy consumption, and has high operational efficiency. Why send computing power to space? - Energy constraints and cooling challenges. With the announcement of large projects like "Stargate" and the continuous increase in Capex by CSP manufacturers, the explosive growth in computing power demand has become a certainty trend, driving a surge in power demand. According to Rands estimates, global AIDC power demand could reach 347 GW by 2030. In terms of cooling, the heat flux density of a 1 million GPU computing cluster currently exceeds 250 W/m^2 locally, necessitating large-scale heat dissipation and cooling measures such as water evaporation. The engineering implementation is extremely difficult, with significant water wastage, and easy to trigger serious heat island effects. For example, Meta's new computing clusters are estimated to use more water than the entire county. Space computing power has advantages in energy, cooling, and other aspects, making it one of the preferred solutions for addressing pain points in the AI era. Is it a fantasy or a certainty? - Projects have been deployed at home and abroad. Currently, multiple companies globally have started corresponding projects. We believe that space computing power technology is not only cutting-edge but also has a large market space and commercial value. For example, the Starcloud project plans to build the first gigawatt-level data center in space; Beijing Xingkongyuan's Orbital Sunrise has completed the first round of financing, the core task is to deploy computing satellites in the Earth's dawn-dusk orbit to form a space data center; China ADA Space and Zhejiang Lab have launched the "Three-Body Computing Constellation" with the first batch of 12 AI satellites launched, capable of performing hundreds of billions of calculations per second, indicating that space computing power may enter the engineering landing phase. In the future, it has strong commercial landing certainty. We believe that based on current technological breakthroughs and industry practices, space computing power has gradually transitioned from concept validation to engineering landing. Space computing power has become a strategic solution to address the challenges of energy and cooling faced by ground computing power. Companies like Nvidia, Amazon, and other AI giants have already started positioning themselves in this trend, and the participation of domestic and foreign giants and policy formulation may accelerate the development of this field, and related investments will begin to emerge. Investment advice: In summary, space computing power already has commercial value potential, and related projects have started to advance with certain methodologies. AI "going to space" is no longer distant, and seizing the opportunity is the key to future success. As a new theme of computing power, it is recommended to pay attention to companies that have laid out in the relevant industry chain in advance, such as US stocks SOCE, RKLB, as well as A shares Shanghai Shunho New Materials Technology, Hang Zhou Iron & Steel, Cetc Potevio Science & Technology, China Spacesat, etc. Risk warning: AI development falls short of expectations, space technology falls short of expectations, high capital requirements, and financing pressure.