马斯克最新宏大设想：每年发射百万吨级的卫星来扩张AI算力

Core Insights - Musk claims that localized AI computing satellites will be the most cost-effective way to generate AI bitstreams within three years and the fastest way to scale within four years [1] - The proposed plan includes launching one million tons of satellites annually, each equipped with 100 kW of power, which could add 100 GW of AI computing power per year without operational or maintenance costs [1] - Musk envisions building satellite factories on the Moon, utilizing mass drivers to accelerate AI satellites to escape velocity without rockets, ultimately aiming for over 100 TW of AI computing power [1] Group 1 - Musk suggests that once factories, robots, and large-scale drivers are established on the Moon, the system could operate independently of traditional currency, using watts and tons as units [3] - The potential for cost-effective deployment and operation of large-scale AI systems in space is highlighted, with predictions that costs could be lower than on Earth within the next 4-5 years due to "free" solar energy and efficient radiation cooling [4] - Musk emphasizes the limitations of Earth, stating that the combined demand for power and cooling from expanding AI computing clusters will exceed the capacity of existing terrestrial infrastructure [4]