China's Three-Body Constellation Has Been Running 8-Billion-Parameter AI Models in Orbit Since 2025

Shanghai has opened its first dedicated space-computing industry hub, in the city's Songjiang district, bringing together aerospace developers, semiconductor manufacturers, AI specialists, cloud vendors, researchers and investors at an inaugural forum on August 31.

The hub arrives more than a year after China's Three-Body Computing Constellation, led by government-affiliated Zhejiang Lab in partnership with Chengdu-based ADA Space, put its first 12 satellites into low Earth orbit in May 2025. Each satellite carries an 8-billion-parameter AI model capable of 744 trillion operations per second, running on commercial off-the-shelf processors inside a compact 6U CubeSat platform.

The satellites communicate with each other using free-space optical links running at up to 100 gigabits per second. On July 29, 2026, the constellation's developers said two satellites had maintained a continuous laser link for 192 hours, 8 minutes and 45 seconds — just over eight days — at 99.99% uptime across distances of up to 1,000 kilometres, with zero data loss.

Li Chao, director of the space-based computing system research center at Zhejiang Lab, said up to 90% of the data satellites generate is currently lost and never effectively processed, because a satellite's downlink window to a ground station may last only about 10 minutes with 1 to 5 gigabits per second of bandwidth. Processing imagery onboard lets a satellite transmit only its conclusions — such as confirmed ship coordinates or weather anomalies — rather than raw data.

The constellation's developers reported a 94% autonomous classification accuracy rate for astronomical and terrestrial infrastructure imagery, a figure that has not been independently audited by Western institutions. They project that at a planned full scale of 2,800 satellites, the network would deliver a combined computing capacity of 1,000 peta-operations per second, linked entirely by laser.