China has launched the first batch of satellites for an ambitious plan to create an artificial intelligence (AI) supercomputer network in space.
On May 14, 12 satellites were sent into orbit aboard a Long March 2D rocket from the Jiuquan Satellite Launch Center. These satellites are the initial components of the planned “Three-Body Computing Constellation” — a 2,800-satellite system developed by ADA Space and Zhejiang Lab.
The goal is to shift data processing from Earth to space, reducing dependence on ground-based computing.
The satellites will use the cold environment of space for natural cooling while performing large-scale data processing at a combined capacity of 1,000 petaFLOPS (1 quintillion operations per second), according to the Chinese government.
Wang Jian, head of Zhejiang Lab, emphasized at the Beyond Expo tech conference in Macau on May 21 that space is becoming a new frontier for innovation in AI.
Typically, satellites send raw data to Earth for processing, but this method is limited by transmission bandwidth and timing constraints. To solve this, China is adopting “edge computing,” where satellites analyze data onboard before sending it back. This not only saves energy—thanks to solar power and natural heat dissipation in space—but also reduces data loss.
Each satellite features an AI model with 8 billion parameters and a processing speed of 744 tera operations per second (TOPS). Collectively, they can reach up to five peta operations per second, significantly outpacing current consumer-grade AI devices like Microsoft’s Copilot+ laptops, which run around 40 TOPS. The satellites are also equipped with laser communication and one includes an X-ray polarization detector to observe high-energy cosmic events.
The constellation is named after the “three-body problem,” a complex physics challenge involving the unpredictable movement of three gravitational bodies, which also inspired Liu Cixin’s sci-fi trilogy The Three-Body Problem. Wang noted that this concept also reflects the complexity of international collaboration, inviting global partners to contribute to and benefit from the system.
Although space-based computing has been tested by the U.S. and Europe, China’s system marks the first full-scale deployment. Meanwhile, former Google CEO Eric Schmidt has also proposed launching orbital data centers, citing the massive energy demands future data centers will face — an estimated 29 gigawatts by 2027 and 67 gigawatts by 2030.
The project signals a major shift toward sustainable and space-based solutions for the growing demands of AI computation.
