China's launch of a 12-satellite cluster on May 14 as part of the "Three-Body Computing Constellation" project is deemed as a crucial step to redefine computing infrastructure by extending it into space. The initiative aims to deploy thousands of satellites, eventually delivering a combined 1,000 peta operations per second (POPS) in computing capacity. 

The 12 satellites operate on a next-generation intelligent satellite platform developed by ADA Space (also known as Guoxing Aerospace) – an Al satellite startup based in Chengdu, southwest China's Sichuan Province. Each satellite is equipped with onboard AI computing systems and inter-satellite communication modules provided by Zhejiang Lab – a research institute based in Hangzhou, east China's Zhejiang Province – forming an interconnected network capable of real-time in-orbit processing. 

The current in-orbit computing capacity of the constellation has reached 5 POPS – comparable to some of the word's top supercomputers, depending on the type of calculations involved. Thanks to laser inter-satellite links transmitting up to 100 Gbps, the constellation can process massive volumes of data directly in space, minimizing the need for Earth-based processing.

"These computing satellites are not only significant for advancing space technologies – such as supporting spacecraft in low-Earth orbit, lunar, and Mars missions – but also open up new possibilities for testing large-scale computing architectures and serving future applications on Earth and in low-altitude airspace," said Zhao Hongjie, executive vice president of ADA Space, in an interview with China Media Group. 

Why is China bringing computing power to space?

With satellite imagery and sensor resolution rapidly advancing, the volume of orbital data is ballooning. "If we continue to transmit all that data back to Earth for processing, we face growing latency and bottlenecks," said Wang Jianyu, an academician at the Chinese Academy of Sciences (CAS) and head of the UCAS Hangzhou Institute for Advanced Study. 

"By processing data in space using AI, we can improve both the efficiency and timeliness of its application across industries," he stated.

These satellites' real-time data processing significantly reduces transmission latency and supports greater autonomy for space systems, especially for autonomous navigation, disaster monitoring, and global AI services.

"The nature of a computing satellite constellation is not about satellites, but computing," said Wang Jian, director of Zhejiang Lab and founder of Alibaba Cloud, during a keynote speech at the BEYOND tech conference in Macao. "Satellites are just a new carrier – much like how personal computers once disrupted mainframes."

Challenges and risks

However, space computing is not without obstacles. The harsh space environment, such as the near-vacuum conditions, high radiation, microgravity, and extreme temperature variations, demands specialized materials and radiation-hardened components to ensure reliability and longevity. 

Cost is another hurdle – with the return on investment timeline remaining uncertain. Launching thousands of satellites requires billions of dollars, on top of the fact that China has yet to master reusable rocket technology to drive down expenses.

In addition, satellites typically have a lifespan of 5 to 10 years, requiring ongoing spending on energy supply, orbital maintenance, and software or hardware upgrades – along with contingency plans for radiation-induced system failures. The price of replacing or upgrading them in orbit would be significant.

Building a space computing infrastructure demands substantial resources and cannot rely solely on a few enterprises. "It requires broad participation and collective input from across society," said Wang Jianyu.

Echoing that view, ADA Space is joining forces with a cohort of 54 universities, research institutions, and enterprises from around the world to jointly launch the "Star Compute" initiative, which aims to build a  2,800-satellite constellation for intelligent space-based computing.