On August 5, 2026, China deployed ten satellites from a floating platform off the coast of Shandong Province. Eight of these units are dedicated to AI computing and laser communications as part of the Star Compute initiative.
The 10-kilowatt cooling system keeping processors under 50°C
The recent launch, utilizing a Jielong-3 carrier rocket, introduced a critical engineering breakthrough for space-based computing. According to AFP, the satellites deployed by ADASpace and a state-run aerospace conglomerate feature 10-kilowatt liquid-cooling systems and deployable thermal radiators. These systems are designed to keep electron temperatures below 50°C even under full-load conditions.
This thermal management is essential because traditional orbital processors often suffer from thermal distortion, which throttles performance. By stabilizing the environment for high-performance computing chips, ADASpace is enabling the satellites to run complex AI workloads—such as real-time data analytics and image classification—autonomously in the harsh vacuum of space.
Building a 2,800-node orbital AI network
The Star Compute programme aims to establish a massive orbital infrastructure consisting of 2,800 computing nodes. This network is intended to provide full spherical coverage of Earth, allowing for high-speed AI processing that supports autonomous vehicle navigation and geospatial anomaly detection. As reported by Chinese state media, this project is a cornerstone of the broader "Digital Silk Road" and "Space Development Plan" initiatives.
Unlike traditioanl satellite constellations, the Star Compute network prioritizes "compute-centric" payloads. This means the satellites do not just relay data; they process it in orbit. To support this, the satellites utilize laser communication prototypes with modulation speeds exceeding 100 Gbps, which far outperform standard radio-frequency links in terms of bandwidth.
Training Alibaba's Qwen-3 model in 48 hours
A primary proof-of-conept for this infrastructure was the recent benchmarking of Alibaba's Qwen-3 language model. The developer announced that the model was trained within a 48-hour runtime while in orbit, a feat that typically requires several days when using terrestrial supercomputers. This suggests that the distributed nature of orbital computing may offer efficiency gains over ground-based data centers.
This capability transforms the satellite from a communication tool into a floating data center. By integrating AI pipelines across a cloud-edge continuum, the Star Compute network could allow millions of Earth-based data points to be processed instantly without the latency of sending data back and forth to a ground station.
A compute-centric alternative to Starlink and OneWeb
The scale of the Star Compute project echoes the mega-constellations deployed by SpaceX's Starlink and OneWeb, but the strategic objective differs. While Western constellations focus primarily on global internet connectivity, China's approach centers on orbital intelligence. This shift follows an earlier deployment in May 2025, where 12 initial satellites established the first horizon-based computing nodes.
This trajectory suggests a new arms race in space, moving beyond who can provide the most bandwidth to who can provide the most processing power. The ability to execute AI inference at the "edge" of space could give the operator a significant advantage in both industrial efficiency and national defense applications .
The cybersecurity risks of distributed orbital training
Despite the technical success, the transition to orbital AI introduces significant vulnerabilities. The Chinese Ministry of Science and Technology has already issued guidelines for secure cryptographic protocols to protect ship-to-satellite and satellite-to-ground data flows, acknowledging that distributed training in space creates a novel attack surface for hackers.
Beyond cybersecurity, international observers remain concerned about the physical impact of 2,800 additional nodes. Specifically, the report leaves several questions unanswered: how will China mitigate the increased risk of space debris, and what specific frequency interference protocols will be used to avoid disrupting existing global communications? Furthermore, the source provides the Chinese government's perspective on national security but does not include commentary from international regulatory bodies regarding the potential militarization of this AI infrastructure.
Comments 0