China is no longer treating orbital computing as a stack of startup pitch decks. On June 3, a new Beijing committee gathered more than 100 applicant organizations around radiation-hardened chips, power systems, thermal management, data transmission, constellations and launch.

The Space Computing Working Committee of the China Computer Industry Association held its inaugural meeting in Beijing under guidance from the Ministry of Industry and Information Technology’s Electronic Information Department, according to SpaceNews. Wang Jianyu of the Chinese Academy of Sciences was elected chairman.

A separate industry summary of the meeting said the committee had already received applications from more than 100 organizations, spanning radiation-hardened chips, space-computing hardware, power and thermal systems, data transmission, constellation infrastructure and launch services, according to Shanghai Metals Market.

Two committees, one Five-Year Plan window

The June body is the second such coordinating committee formed in 2026. The first, the Space Computing Power Professional Committee, was stood up earlier in the year to focus more heavily on standards, applications and terrestrial-space integration.

The new committee tilts toward hardware and supply chain coordination. That distinction matters because the bottlenecks are physical: chips that survive radiation, systems that reject heat, lasers that move data between spacecraft, and launch cadence high enough to make the whole network more than a demonstration.

The timing is not accidental. China approved the outline of its 15th Five-Year Plan in March 2026, setting the national development window for 2026 through 2030, according to the State Council.

The plan’s space sections emphasize a space-air-ground network that combines communications, internet connectivity, positioning and remote sensing, according to CGTN’s summary of the blueprint. Space computing fits into that architecture as the layer that decides what data moves, what stays in orbit, and what gets processed before it ever touches a ground station.

That is the real industrial bet. Not simply putting processors in orbit, but making orbit part of the data infrastructure itself.

The companies trying to put processors above the atmosphere

Beneath the committees, several Chinese firms are already pushing hardware toward flight. Beijing-backed Orbital Chenguang secured $8.4 billion in credit lines in April, a figure that signals state-aligned banking support rather than ordinary venture appetite, according to SpaceNews.

Chinese project-finance summaries describe Orbital Chenguang’s target as a phased orbital data-center buildout that moves from early computing satellites toward larger low-Earth-orbit infrastructure after 2030. That is a long way from a working gigawatt-class orbital data center.

Shanghai Bailing Aerospace is working at a smaller scale. Public reports describe early-stage funding of tens of millions of yuan and a plan to develop a 100-kilowatt-class in-orbit computing platform.

Shanghai-based Oriental Tiansuan and photonics startup Guangbenwei have also announced work on a space-based optical computing satellite. The claim is still in announcement territory, but it shows how broad the technical sweep has become: electronic compute, optical compute, edge processing, inter-satellite links and power systems are all being pulled into the same policy frame.

Space Daily has covered the same market split outside China, where orbital data-center companies are dividing the stack into power, hardware and application layers. China’s version is different because the state is trying to coordinate the layers at once.

The most concrete hardware is already in orbit

The clearest demonstration so far is the Three-Body Computing Constellation. On May 14, 2025, China launched 12 satellites on a Long March 2D rocket from Jiuquan as the first batch of the constellation, according to the State Council’s English portal.

The mission is led by Zhejiang Lab and is intended to grow into a network of thousands of satellites with total computing power of 1,000 peta operations per second, or 1,000 POPS. That is an exascale-class number, though it is not the same metric as the floating-point LINPACK figures used to rank conventional ground supercomputers.

The first 12 satellites were reported to have a combined computing capacity of 5 POPS and inter-satellite laser links. They are designed to process data in orbit instead of sending every raw stream back to Earth first.

In January 2026, Xinhua reported that GuoXing Aerospace had deployed a general-purpose AI model aboard orbiting satellites and run experiments in which prompts were sent from Earth, processed on board and returned to ground stations in roughly two minutes, according to the National Center for Science and Technology Infrastructure.

That makes the Chinese program more than a paper exercise. It also keeps the scale in perspective: 12 satellites and 5 POPS are real hardware, while thousands of satellites and 1,000 POPS remain a target.

Why orbital computing is tempting and unforgiving

The case for processing data in orbit begins with bandwidth. Earth-observation satellites, radar satellites and sensor platforms can collect more raw information than they can conveniently downlink, especially when ground-station windows are short or contested.

If a satellite can sort images, filter noise, detect ships, flag wildfires or compress sensor returns before transmission, the bottleneck changes. The spacecraft sends the useful result instead of the entire raw stream.

Solar power is another attraction. In the right orbit, solar panels can see long, predictable stretches of sunlight without clouds, land constraints or local grid interconnection queues.

Cooling is more complicated than the pitch decks make it sound. Space is cold in the popular imagination, but a satellite can shed heat only by radiating it away, and radiator area becomes a punishing design constraint as power density rises.

The same physical math sits behind the global launch debate. Space Daily has previously written about why cheaper launch does not automatically make more satellites a cleaner or simpler system, and orbital data centers would push that question harder than communications constellations do.

The watch points are hardware, heat and standards

The first thing to watch is whether Orbital Chenguang draws on its credit lines for actual flight hardware. A credit facility is not a deployed satellite, and it is not a functioning orbital data center.

The second is whether the Three-Body constellation moves from 12 satellites toward a meaningful fraction of its planned scale before 2030. The difference between a demonstration cluster and a thousand-satellite compute network is not incremental.

The third is standards. If the committees begin publishing technical interfaces for orbital compute modules, radiation-hardened processors, optical inter-satellite links or space-ground data routing, that would be a more durable sign of industrial coordination than another funding announcement.

The hardest question is still the chip and thermal stack. Radiation-tolerant processors have to survive the environment, produce enough useful work, reject heat through spacecraft radiators and still beat the economics of ground infrastructure.

China has built a policy frame around that question. The frame is visible now: committees, Five-Year Plan language, state-aligned credit, launch capacity and a first batch of AI-capable satellites already circling Earth.

The machines themselves still have to prove the frame can hold. Above the atmosphere, the data can move only through power, heat, light and orbit, and every one of those constraints is less forgiving than a planning document.