The Shift to Orbital Infrastructure: Beyond Ground-Based AI
For decades, the heavy lifting of artificial intelligence has happened in massive, energy-hungry data centers on Earth. Yet, a fundamental shift is occurring as the industry looks toward the stars. The emergence of space-based computing infrastructure aims to move data processing from the ground to the orbit, creating a global network of “supercomputers in the sky.”
Leading this charge is the “star compute” initiative by ADAspace. This ambitious vision involves the deployment of 2,800 computing satellites—comprising 2,400 inference computing satellites and 400 training computing satellites. Positioned at altitudes between 500 and 1,000 kilometers in sun-synchronous, dawn-dusk, and low-inclination orbits, this network is designed to deliver inference power at the hundred-thousand-petaflop level and training power at the million-petaflop level.
Solving the Ground-Based Computing Bottleneck
The push for orbital computing isn’t just about exploration; it is a practical solution to the limitations of terrestrial infrastructure. Ground-based data centers are facing a critical bottleneck, particularly regarding energy and accessibility.
Current global data centers consume approximately 1.5 percent of the world’s total electricity generation, with 40 percent of that energy dedicated solely to cooling. Vast portions of the planet remain “network blind zones,” including 70 percent of the world’s land and 95 percent of the oceans, where transoceanic fiber optics cannot support real-time interactive applications.
The Orbital Advantage
Space offers two primary advantages that ground centers cannot match:
- Energy Efficiency: Space-based solar power generation is several times more efficient than ground-based photovoltaics and remains virtually unaffected by weather or day-night cycles.
- Natural Cooling: The deep-cold environment of space, which approaches absolute zero, can significantly reduce the energy required for cooling high-performance chips.
To facilitate these launches, reliable carrier rockets like the Long March 2D have been utilized to deploy these critical constellations into orbit.
From Theory to Product: Real-World Applications
Space computing is transitioning from a technical experiment to a commercial product. This is evidenced by the launch of “Prometheus,” the world’s first space-based computing power cloud service platform for enterprises.
Recent technical breakthroughs demonstrate the tangible impact of this technology:
- Robotic Control: In collaboration with Shanghai Jiao Tong University, ADAspace successfully used space-based computing power to control a ground robot. This suggests a future where humanoid robots, autonomous vehicles, and drones can operate without blind spots, calling upon orbital computing when ground networks are unavailable.
- In-Orbit Large Language Models (LLMs): The deployment of Alibaba’s Qwen3 large language model to a satellite cluster marked the first in-orbit deployment of a general-purpose large model, with end-to-end inference completed in less than two minutes.
- The Three-Body Constellation: Zhejiang Lab is pursuing the “Three-Body Computing Constellation,” aiming for a scale of 100 satellites by 2027 to achieve advanced in-orbit data processing.
Overcoming the Hostility of the Vacuum
Despite the potential, building a computer that survives in space is an immense engineering challenge. Hardware must be designed to withstand extreme vacuum, intense radiation, and drastic temperature fluctuations.
The China Academy of Information and Communications Technology (CAICT) has established a professional committee to tackle these hurdles. Their research focuses on four key pillars:
- Space-borne AI chips: Developing radiation-resistant processors.
- Inter-satellite laser communication: Ensuring high-speed data transfer between nodes.
- Efficient thermal control: Managing heat dissipation in a vacuum.
- Space photovoltaics: Maximizing energy capture.
The Economic Horizon of Space Computing
The financial implications of this industry are vast. According to preliminary estimates from the CAICT, the scale of China’s space computing power industry is expected to exceed 250 billion yuan (approximately 36.6 billion U.S. Dollars) by 2030.
With the Ministry of Industry and Information Technology supporting the cultivation of a comprehensive ecosystem—including standards for hardware, software, and security—space-based computing is poised to become a universal public service. Much like water or electricity, orbital computing power could soon provide the backbone for deep-sea exploration, drone swarm coordination, and autonomous global transport.
For more insights into the evolution of orbital tech, explore our guides on emerging space technologies and AI infrastructure evolution.
Frequently Asked Questions
What is space-based computing?
It is the practice of deploying computing hardware (like AI chips and servers) on satellites to process data in orbit rather than sending all raw data back to Earth for processing.
Why is it better than ground-based computing?
It eliminates network blind zones (covering oceans and remote land), utilizes more efficient solar power, and leverages the cold of space for natural cooling.
What is the “Star Compute” plan?
An initiative by ADAspace to build an AI space infrastructure of 2,800 satellites to provide global training and inference computing power.
What are the main challenges of space computing?
The primary obstacles include extreme radiation, temperature swings, the difficulty of hardware maintenance after launch, and the need for specialized radiation-resistant chips.
What do you think? Will orbital computing eventually replace terrestrial data centers for AI, or will it remain a niche tool for remote operations? Share your thoughts in the comments below or subscribe to our newsletter for the latest updates on the space economy!
