Feb 7, 2026

China's Three-Body Computing Constellation: 2,800 Satellites for Orbital AI Supercomputing

China is building the world’s first operational orbital supercomputer network. The Three-Body Computing Constellation, named after Liu Cixin’s science fiction novel, launched its first 12 satellites on May 14, 2025, from Jiuquan Satellite Launch Center. The project targets 2,800 satellites by 2030, creating a space-based AI infrastructure with 1,000 peta operations per second (POPS), one quintillion operations per second.

This is not a proposal. It is operational hardware in orbit right now.

Current Status: 12 Satellites, 5 POPS

The initial cluster, launched aboard a Long March 2D rocket, is led by ADA Space and Zhejiang Lab. Each satellite performs 744 trillion operations per second. The 12-satellite cluster achieves a combined 5 POPS, roughly equivalent to mid-tier terrestrial supercomputers.

The satellites process data in orbit rather than transmitting raw data to ground stations. This architecture reduces bandwidth requirements and enables real-time applications that cannot tolerate ground-to-space-to-ground latency.

Technology Readiness Level: TRL 5-6 (operational demonstration in relevant environment)

Why Space-Based Computing Works

Orbital data centers leverage the vacuum of space for passive cooling. Terrestrial data centers consume massive energy for thermal management. In orbit, radiative cooling is free. No air conditioning, no water cooling loops, no energy wasted fighting thermodynamics.

China’s 15th Five-Year Plan (2026-2030) designates the establishment of the first space-based AI data center as a national goal. The government expects operational capacity within five years from February 2026.

The Three-Body Constellation targets applications where orbital computing provides clear advantages:

Disaster detection: Real-time analysis of satellite imagery without ground relay delays
Climate monitoring: Continuous atmospheric data processing at the collection point
Deep-space observation: On-orbit processing of astronomical data from space telescopes
Military tracking: Real-time surveillance without ground station dependencies

The 2,800-Satellite Architecture

Full deployment occurs “over the next several years,” according to official statements. The complete constellation will distribute 1,000 POPS across 2,800 satellites, averaging 357 trillion operations per second per satellite.

For comparison, the current top terrestrial supercomputer, Frontier at Oak Ridge National Laboratory, achieves 1.2 exaflops (1,200 POPS). China’s orbital network would reach 83% of Frontier’s performance, distributed across thousands of nodes in low Earth orbit.

This is not theoretical. The first 12 satellites are operational. The launch cadence and satellite production capacity determine the timeline, not the technology.

Comparison to Other Orbital Computing Projects

China is not alone in pursuing space-based computing, but it is ahead in operational deployment:

SpaceX Orbital Data Center Filing (January 2026): Proposed 1 million satellites for 100 GW annual AI compute capacity. TRL 2-3 (concept with initial engineering).
Google Project Suncatcher: Tensor processing units on satellites by 2027. TRL 3-4 (prototypes under development).
ESA ASCEND: European orbital data center demonstration mission planned for 2026. TRL 3-4 (demonstration mission planned).

China’s Three-Body Constellation is the only project with operational satellites performing orbital computing right now.

Challenges and Limitations

Space-based computing faces real constraints:

Radiation hardening: Commercial processors fail in space radiation environments. Radiation-tolerant designs sacrifice performance. The Three-Body satellites use unspecified processors. Performance per watt and radiation tolerance are not publicly documented.

Thermal cycling: Satellites experience extreme temperature swings between sunlight and shadow. Thermal stress degrades electronics over time.

Latency: Orbital computing eliminates ground-to-space data transmission latency but introduces inter-satellite communication delays. Applications requiring sub-millisecond response times cannot use orbital processing.

Debris proliferation: 2,800 satellites add to the growing orbital debris problem. End-of-life deorbiting plans are not publicly detailed.

Path Forward

The Three-Body Computing Constellation validates the orbital computing model. Twelve satellites are operational. The technology works. The question is scalability: can China manufacture and launch 2,788 more satellites while maintaining reliability, managing thermal and radiation challenges, and preventing debris accumulation?

The 15th Five-Year Plan provides the answer timeline. By 2030, we will know if orbital supercomputing is a viable alternative to terrestrial data centers or a niche application for specific use cases.

The first 12 satellites prove the concept. The next 2,788 determine the future.