According to the information from Zhitong Finance APP, Guosheng Securities published a research report stating that with the announcement of large-scale project plans such as ‘Stargate’ and the continuous upward revision of Capex by various CSP vendors, the explosive growth in computing power demand has become a certainty. However, ground-based infrastructure may face bottlenecks related to energy, heat dissipation, and land. The report points out that space-based computing power has emerged as a strategic solution to address the energy and heat dissipation challenges faced by terrestrial computing. Initial moves by AI giants like NVIDIA and Amazon indicate this trend, and the participation of global leaders alongside supportive policies could accelerate development in this field, with relevant investments beginning to materialize.
The main points from Guosheng Securities are as follows:
At this juncture, we continue to explore other potential extensions of AI infrastructure. While the market is focused on hardware facilities within computational infrastructure, such as optical modules and switches, or eagerly anticipating breakthroughs in AI applications, the underlying logic regarding energy, space, and heat dissipation may undergo fundamental changes when computational scale surpasses critical thresholds. As data center electricity demands surge and American electric infrastructure lags behind, projects like Starcloud’s gigawatt-level orbital power stations and China’s ‘Three-Body Computing Constellation’ are demonstrating real-time responsiveness: the computational battleground may have already extended from the ground to outer space.
What is space-based computing power? It refers to data centers deployed in Earth’s orbit. Space-based computing involves deploying data centers and computational capabilities into space through satellites equipped with computing hardware for on-orbit data processing. Utilizing high-speed laser communication between satellites enables data transmission and real-time processing, with results sent back to Earth. Due to unique conditions in space, such as a vacuum environment and specific lighting conditions, space-based computing exhibits characteristics of autonomous intelligence, real-time responsiveness, distributed collaboration, minimal energy consumption, and high computational efficiency.
Why move computing power to space? — Energy constraints and heat dissipation challenges. With the announcement of large-scale initiatives like ‘Stargate’ and the continued upward revision of Capex by various CSP vendors, the explosive growth in computing power demand has become an established trend, driving a sharp increase in electricity needs. According to Rand estimates, global AIDC electricity demand will reach 347GW by 2030. In terms of heat dissipation, current GPU clusters of one million units generate localized heat flux densities exceeding 250W/㎡, necessitating extensive heat diffusion measures and cooling techniques such as water vapor evaporation. Such engineering implementations are highly complex, waste significant water resources, and can easily lead to severe urban heat island effects—as seen with Meta’s new computing cluster expected to consume more water than the entire county’s usage. Conversely, space-based computing offers advantages in energy and heat dissipation, positioning it as a preferred solution to these pain points in the AI era.
Is this concept far-fetched or does it hold tangible prospects? — Projects are already underway globally. Several companies worldwide have initiated corresponding projects, and we believe that space-based computing technology is not only cutting-edge but also holds substantial market potential and commercial viability. For instance, Starcloud plans to construct the first gigawatt-level data center in space; Beijing Starlight Institute’s Orbital Dawn has completed its initial financing round, with its core mission being the deployment of computing satellites in Earth’s twilight orbit to form a space-based data center. Additionally, China’s ADA Space, in collaboration with Zhejiang Lab, launched the first batch of 12 AI satellites under the ‘Three-Body Computing Constellation,’ with expected computing power reaching exascale operations per second—indicating that space-based computing may be entering an engineering implementation phase with strong prospects for commercial success.
We believe that based on current technological breakthroughs and industry practices, space-based computing has transitioned from proof-of-concept to an engineering implementation stage. Space-based computing has emerged as a strategic solution to resolve the energy and heat dissipation issues associated with terrestrial computing. The initial moves by AI giants like NVIDIA and Amazon underscore this trend, while participation from global leaders and policy support may further accelerate progress in this area, making relevant investments increasingly visible.
Investment Advice: In summary, space-based computing holds significant commercial value potential, with relevant projects underway and methodological support. The idea of AI going ‘space-bound’ is no longer far-fetched, and seizing early opportunities will be key to future success. As a new frontier in computational capacity, we recommend paying attention to companies that are strategically positioned in this sector, such as SOCE and RKLB in the US stock market, and A-share companies like Shunhao Co., Ltd., Hangzhou Steel Co., Ltd., Potevio Technology, and China Satellite.
Risk Warning: AI development falling short of expectations, space technology falling short of expectations, high capital requirements, and financing pressures.