With the soaring demand for AI computing capabilities, the energy consumption challenges faced by terrestrial data centers have become ever more apparent. Tech behemoths in Silicon Valley are now venturing into space-based data centers as a novel frontier in the fiercely competitive AI landscape. Jensen Huang, CEO of NVIDIA, recognizes the immense long-term potential of orbital computing. However, he also underscores the myriad obstacles in setting up AI data centers in space, with cooling systems emerging as a particularly vexing issue. Given that space lacks the ability to dissipate heat through conduction and convection, relying solely on radiation necessitates the use of large-scale heat dissipation mechanisms. Overcoming this challenge may well take several years. Nevertheless, NVIDIA has already embarked on preliminary explorations by deploying CUDA-based systems on satellites and launching the Space-1 Vera Rubin module. The module's THOR chip has successfully cleared 'radiation certification' and is currently employed for image processing tasks on satellites. Huang is optimistic that the economic advantages of space data centers will progressively become more evident.