4.1 KiB
| type | title | author | url | date | domain | secondary_domains | format | status | priority | tags | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| source | China Completes In-Orbit Testing of Three-Body AI Computing Constellation | SatNews Staff (@SatNews) | https://satnews.com/2026/02/16/china-completes-in-orbit-testing-of-three-body-ai-computing-constellation/ | 2026-02-16 | space-development |
|
article | unprocessed | high |
|
Content
China has successfully concluded nine months of orbital testing for its "Three-Body" Computing Constellation, developed by ADA Space and Zhejiang Lab (with CASIC involvement). The constellation demonstrates the ability to run large-scale artificial intelligence models directly on satellite hardware, drastically reducing the latency between data capture and actionable intelligence.
Technical specifications confirmed:
- 12 satellites launched May 14, 2025 on Long March 2D from Jiuquan
- 744 TOPS per satellite; ~5 PFLOPS collectively
- 100 Gbps laser inter-satellite links; 30 TB on-orbit storage
- Running an 8-billion-parameter remote sensing LLM and an 8-billion-parameter astronomical time-domain model — among the largest parameter counts of any in-orbit AI globally
- 94% classification accuracy without ground intervention
- Satellites demonstrated real-time data sharing between units for distributed computing
Expansion plan:
- 32-satellite "Computing Grid" by 2028
- 2,800-satellite "Star-Compute Program" total
- 1,000+ POPS target at full constellation
Agent Notes
Why this matters: China's Three-Body constellation is the world's most advanced operational orbital AI computing system. It is operationally ahead of comparable US programs. This is not speculative or conceptual — 9 months of in-orbit validation is complete. The KB had no claim capturing this milestone.
What surprised me: Previous sessions (including 2026-04-22) treated the "Three-Body" references in military-framing SpaceNews pieces as possibly conceptual/unconfirmed. It is definitively real, operational, and CIVILIAN/commercial (ADA Space + Zhejiang Lab), not purely military. This requires correcting the prior session's characterization.
What I expected but didn't find: Any US equivalent with comparable in-orbit validation. The US ODC market has operators running production workloads (Kepler, others) but none with 9 months of systematic in-orbit testing at this scale.
KB connections:
- Relates to existing KB claims about orbital data centers being gated on launch costs — this challenges the threshold framing for the captive compute use case
- Relates to Belief 7 (single-player SpaceX dependency) — China is building parallel infrastructure that functions independently of Western launch
Extraction hints:
- Primary claim: China's Three-Body Computing Constellation is the world's most operationally advanced orbital AI system as of February 2026
- Secondary claim: China's orbital computing is civilian/commercial-led (ADA Space/Zhejiang Lab) not purely military, challenging the "military-first" narrative
- Possible divergence: KB may have claims about orbital computing being speculative/pre-commercial that contradict this
Context: This is published by SatNews, a reliable space industry trade publication. ADA Space is a Chinese space startup; Zhejiang Lab is a major Chinese AI research institution (similar to a national AI lab). CASIC (China Aerospace Science and Industry Corporation) is the state defense contractor involved.
Curator Notes (structured handoff for extractor)
PRIMARY CONNECTION: Claims about orbital data center market being gated on launch cost thresholds — this source provides counter-evidence for the captive compute segment WHY ARCHIVED: China's operational superiority in orbital AI computing as of February 2026 is a fact the KB needs to reflect; also needed to correct the military-framing characterization in the archived Armagno/Crider piece EXTRACTION HINT: Focus on two distinct claims: (1) Three-Body is operational not speculative, (2) civilian/commercial operator (not just military), contradicting prior characterization