STAR.VISION Plans 1,080-Satellite Space.IDC Compute Constellation

The project is called the Space.IDC Computing Constellation. China in Space reported on 29 September that the company announced it at the Global Digital Trade Expo on 25 September. It joins a quiet race among Chinese firms to put data centre capacity into orbit.

The company says about 1,080 spacecraft will make up the main constellation, fully deployed by 2035. Roughly 720 will handle inference tasks and 360 general compute. At the unveiling the company gave no orbit details, no inter-satellite link specifications, no radiator sizes and no mass figures, and it has not yet filed international regulatory paperwork, the outlet says. STAR.VISION has published three spacecraft designs. The G1 verification craft is meant to deliver 40 to 60 PetaFLOPs from 11.5 kilowatts of solar generation. The G2 standard design is quoted at 160 to 310 PetaFLOPs and 48 kilowatts. The G3 flagship is listed at 460 to 1,000 PetaFLOPs and 140 kilowatts. Launches for the G1 units have been booked on Orienspace's Gravity-1 and Galactic Energy's Ceres-2, according to the same report. No target month or quarter was given. The silence on orbits and cooling is not a small omission. A constellation of this size radiates waste heat in vacuum and moves data between spacecraft, and both of those jobs depend on exactly the two things the company left out.

There are numbers, but not enough of them

What the company did disclose is money. Strategic credit agreements with the Industrial and Commercial Bank of China and the Agricultural Bank of China total 10 billion Yuan, or 1.49 billion dollars as of 28 September. At least 1 billion Yuan (149 million dollars) is flagged for Phase One. The company did not share further terms. Partners named include the Hangzhou Institute for Advanced Study at UCAS, the Hangzhou Research Institute of Xidian University, Hongqing Technology, Baidu AI Cloud and MinoSpace. Government bodies in Zhejiang province, Xichang in Sichuan and Chongzhou in Guangxi provide unspecified support.

A crowded field, and the state is watching

STAR.VISION is not first. China in Space notes that ADA Space and Zhejiang Lab's Three-Body Computing Constellation is ahead of the others, with a dozen satellites already operating. Orbital Chenguang has heavier financial backing, and Chaozhisuan Beijing Technology has well-connected partners; both have demonstration satellites in orbit. State-owned firms are also working on large computing spacecraft with far less publicity. The outlet also reports that the central government has designated the relevant technologies as a focus area for the 15th Five-Year Plan period, 2026 to 2030, with the Ministry of Industry and Information Technology involved. That framing matters more than any single launch contract. It suggests Beijing now treats orbital compute as infrastructure policy, not a startup experiment.

Chinese firms are not alone in pushing intelligence toward the edge of the network. IEEE Spectrum reported on 29 September that NASA's Jet Propulsion Laboratory used Anthropic's Claude models last December to help plan two Mars drives for the Perseverance rover, with human planners checking the route before upload. In May, NASA and IBM put a compressed AI model on the International Space Station and a satellite to identify floods and clouds from orbit. In July, astronauts on the ISS tested a large language model for maintenance questions. Those experiments point at the same argument STAR.VISION is making: as missions grow more distant and more numerous, sending everything back to Earth for processing gets slower and more expensive. IEEE Spectrum quotes Robert Ambrose, former chief of NASA's Software, Robotics and Simulation Division, on why engineers have resisted this for decades.

Autonomy often does not have a deterministic outcome, which means that how you got into a certain situation changes the behavior. So if you come into the same situation but from different paths, the outcome could be different. And so engineers hate that.

The gap between a compute constellation and an autonomous spacecraft is worth keeping in view. STAR.VISION's pitch is capacity: raw processing and inference for customers who need work done within China or from orbit. On the disclosed record, it is not a claim that satellites will decide for themselves what to do. Orbital compute and orbital autonomy are related problems, but they are not the same problem, and the company has published numbers for only one of them.

That leaves the usual set of unanswered questions. Cost per PetaFLOP in orbit, thermal management at 140 kilowatts, data downlink, and whether customers will pay a premium for compute that cannot be serviced after launch. None of these appeared in the 25 September announcement. The credit lines are real and the launch bookings are real, at least as announced. The constellation itself is a decade-long plan with a first phase that has not been priced in public. Readers should treat the 2035 deployment date as a target, not a schedule.