China plans orbital data centers with million-card clusters
On November 27, 2025, Beijing's Science and Technology Commission and the Zhongguancun Science City Management Committee released a strategic plan to build large-scale space data centers in low Earth orbit. The centers will operate at altitudes of 700-800 kilometers (435-500 miles) and deliver over 1 GW of computing power per unit, supporting million-card server clusters.
The system will consist of three subsystems: orbital computing, relay transmission, and ground control. The orbital segment will host multiple data centers, each capable of real-time data relay and edge computing services.
Construction will proceed in three phases:
• 2025-2027: Breakthroughs in energy and heat dissipation, launch of test satellites, and deployment of a 200 kW computing constellation with 1,000 POPS.
• 2028-2030: In-orbit assembly and cost reduction, expansion to second-phase constellation, enabling "Earth-Data-Sky" integration.
• 2031-2035: Mass satellite production and grid deployment of orbital data centers for "space-based master computing."
Compared to terrestrial data centers, orbital systems offer natural advantages in energy and cooling. Space provides abundant solar power and low ambient temperatures, eliminating land use and power grid constraints while solving heat dissipation challenges.
The innovation consortium has already developed the first-generation test satellite, Chenguang-1, which is undergoing final assembly and is scheduled for launch by early 2026.
The space data center initiative is led by Beijing's municipal science authorities and the Zhongguancun Science City. Chenguang-1 was developed by a consortium of aerospace and computing firms. Each orbital data center is expected to weigh several tons and span over 10 meters (33 feet) in length. Development costs are reported in Chinese yuan (CNY) and converted to U.S. dollars (USD); a single unit is estimated at ¥500 million (about $70 million USD). The system is designed to support AI, satellite internet, and national data infrastructure.
