I recently came across a set of data, and it's quite interesting.

SpaceX's Starlink project now has 9,500 satellites in orbit, covering 155 countries and regions, serving over 9 million users. Just on January 4th, they launched 29 new satellites using a brand new Falcon 9 rocket, completing the launch and recovery in one go.

Elon Musk has always been a man of his word. Reusable rockets are a reality, and global satellite internet has become a reality. If he now says the annual production of a thousand Starships, who dares to laugh?

But Starlink itself is not the main point of today's discussion. What I am more concerned about is a logic hidden behind Starlink — since satellites can provide internet services, can they also be used to provide computing power services?

Currently, a major pain point in the AI industry is energy. 40% of data center electricity costs are spent on cooling. Industry forecasts predict that by 2027, 40% of data centers will shut down due to lack of power. Problems that cannot be solved on Earth might be solvable in space.

The advantages of space are very hardcore. Infinite solar energy and the natural cooling environment of -270°C in space. Sending AI servers into space means near-zero electricity costs and zero cooling costs.

Starlink has already proven that large-scale satellite networks are entirely feasible. The next level of imagination is to build a space computing network. Evolving from Starlink 1.0 to the 2.0 era of space computing. This idea sounds completely different.