In recent years, the notion of 'space data centers' has drawn considerable interest. However, the actual number of GPUs deployed in space orbit is still notably limited. Presently, this scenario is undergoing a gradual transformation, and the business model for orbital computing power is poised to become more distinct in the foreseeable future. In November 2025, Starcloud successfully launched the world’s inaugural space data center test satellite, which is armed with an NVIDIA H100 GPU. This satellite boasts an impressive computing power of 2000 TFLOPS on a single unit. At the 2026 GTC conference, NVIDIA unveiled the Space-1 Vera Rubin module, which marks a significant leap in space AI inference computing power, achieving a performance boost of up to 25 times compared to the H100. Both domestic and international tech firms are fervently pushing forward with the construction of space data centers. Their goal is to slash construction costs through modular and standardized designs, paving the way for potential large-scale networking applications in the future.