China Launches AI-Powered Satellites to Supercharge Space Computing

On May 14, China launched a cluster of 12 satellites as part of its ambitious "Three-Body Computing Constellation" project—an effort to shift computing infrastructure from Earth to orbit. Imagine data centers floating in space, processing information at lightning speed without the usual latency of ground-based systems.

Building a Space-Based Supercomputer

The satellites, built by ADA Space (also known as Guoxing Aerospace) in Chengdu, Sichuan Province of the Chinese mainland, each carry advanced AI computing modules and inter-satellite links developed by Zhejiang Lab in Hangzhou, Zhejiang Province of the Chinese mainland. Together, this first batch delivers 5 peta operations per second (POPS)—comparable to some of today’s top supercomputers.

Thanks to laser links capable of 100 Gbps, the constellation can process massive volumes of imagery and sensor data directly in orbit, reducing the need to transmit every bit back to ground stations.

Wang Jian, director of Zhejiang Lab and founder of Alibaba Cloud, framed the shift at the BEYOND tech conference in Macao: “The nature of a computing satellite constellation is not about satellites, but computing. Satellites are just a new carrier—much like how personal computers once disrupted mainframes.”

Why Move Computing to Space?

As satellite imagery resolution and sensor capabilities soar, so does the volume of data. “If we continue to transmit all that data back to Earth for processing, we face growing latency and bottlenecks,” says Wang Jianyu, academician at the Chinese Academy of Sciences and head of the UCAS Hangzhou Institute for Advanced Study.

“These computing satellites are not only significant for advancing space missions in low-Earth orbit, the Moon, and Mars—they also offer new opportunities for testing large-scale architectures and serving Earth-based needs,” adds Zhao Hongjie, executive vice president of ADA Space.

Scaling Up—and Facing Challenges

China’s goal is to eventually field thousands of satellites, delivering up to 1,000 POPS of computing capacity. However, space is unforgiving: near-vacuum conditions, radiation, microgravity, and extreme temperatures demand specialized materials and radiation-hardened components.

Cost is another barrier. Without mature reusable rocket technology, launching and maintaining a massive constellation will require billions of dollars and ongoing investments in power, orbital upkeep, and hardware upgrades. Satellites typically last 5–10 years, and replacing them in orbit is no small feat.

“Building a space computing infrastructure cannot rely on a few enterprises alone,” warns Wang Jianyu. “It requires broad participation and collective input from across society.”

The Road Ahead: Star Compute

To meet that need, ADA Space has teamed up with 54 universities, research institutes, and enterprises worldwide in the “Star Compute” initiative. This coalition aims to launch a 2,800-satellite constellation, pooling resources and expertise to revolutionize AI computing in space.

Whether it’s speeding up Earth observation, supporting lunar exploration, or unlocking new AI-driven services, this orbiting data center could mark a new chapter in computing history. What breakthroughs might emerge when our data centers leave the ground?

Reference(s):

From Earth to orbit: China's ambition to advance AI space computing

cgtn.com