【#Tech24H】As Earth observation resolution continues to improve, the volume of data satellites send back every day has reached astronomical levels. According to industry statistics, China acquires more than 20 TB of raw remote sensing data per day; a new-generation ultra-wide-swath remote sensing satellite has a data throughput of nearly 50 GB per second during imaging, roughly equivalent to the data volume of a two-hour 4K high-definition movie. However, observation satellites are not always connected to the ground: ground stations are mainly built within China, and satellites can only transmit data during the brief few minutes when they pass overhead, while under natural conditions the interval between two transmission windows is often about ten days. The fact that satellites can capture data but not use it in time has become a major bottleneck for the space-air information industry. Data generated in space is processed in space, an approach the industry calls “space data, space computing”. In the short term, the most valuable applications are those tasks that must be computed in space first and have very high time value, such as emergency response, national defense, and real-time monitoring. As launch costs, in-orbit power consumption, and computing costs continue to decline, space-based computing power is gradually becoming a resource that can be called upon on demand, and may in the future be directly accessed by ordinary industries and even individuals, much like cloud computing. By then, the significance of space computing power will no longer be limited to serving satellites themselves, but will become a new type of digital infrastructure connecting real-world data with artificial intelligence.[ By Zhang Liyan | Tang Ruohan ]