China’s First AI Satellite Cluster Takes Supercomputing to Space

China has successfully launched the first cluster of 12 satellites of its ambitious Three-body computing constellation on May 14 from the Jiuquan Satellite Launch Center aboard a Long March 2D rocket. The complete supercomputing network will comprise 2,800 satellites and will deliver a combined computing capacity of 1,000 peta operations per second (POPS). The twelve satellites currently in orbit are interconnected through high-speed laser communication links that transfer data at up to 100 gigabits per second. This initiative will be using the cold vacuum as a natural cooling system and it might be able to shift the AI processing from Earth to the space. This project is led by the company ADA Space and Zhejiang Lab.

AI in Orbit

According to a translated ADA Space statement, each satellite in China’s launch contains an 8 billion-parameter AI model that can perform 744 tera operations per second (TOPS). When combined together, their processing power will be able to perform five peta operations per second.
Traditionally, satellites collect raw data and transmit it back to Earth, limited by narrow bandwidths and transmission windows. This causes data loss and processing delays. By processing data in space, the system cuts transmission needs and takes advantage of solar power and space’s natural cooling, reducing energy consumption and environmental impact.

Global Cooperation and Cosmic Inspiration

According to Zhejiang Lab’s Wang Jian, the platform is open to international use, encouraging other countries and organizations to take part in its development. Although the U.S. and Europe have performed tests on space computers, China’s array is the first to be deployed at an operable scale. Meanwhile, former Google CEO Eric Schmidt, following his purchase of a controlling interest in the California launch company Relativity Space, has proposed launching data centers into orbit.