Space-Based AI Data Centers Drive Global Competition

Efforts to build AI data centers in space are intensifying, with government agencies and private companies in China and the US launching new initiatives and pilot projects to develop orbital computing infrastructure powered by solar energy.

China Aerospace Science and Technology Corporation has announced a five-year plan to establish gigawatt-class digital intelligence infrastructure in orbit, aiming to integrate cloud, edge, and device-level computing for processing Earth-collected data in space. A December policy document outlines China’s objective to create an industrial-scale “Space Cloud” by 2030, combining computing, storage, and transmission capabilities using solar power as part of the country’s 15th Five-Year Plan.

In the United States, SpaceX has stated plans to deploy solar-powered AI data center satellites within the next two to three years. Elon Musk said at the World Economic Forum that space offers uninterrupted solar power and no night cycles, which he argued makes it the lowest-cost location for AI operations. SpaceX reportedly intends to allocate funds from a planned $25 billion IPO to support the development of these orbital AI systems.

Several pilot projects have already been launched. In May 2025, Chinese startup Adaspace placed 12 “AI-Cloud” satellites into orbit, which are designed to form part of a larger constellation capable of running advanced AI models and transmitting data at high speeds. Meanwhile, a Beijing research institute has published plans to build 16 centralized space data centers in dawn-dusk orbit, with phased development scheduled through 2035.

What the numbers show

• China’s “Space Cloud” aims for completion by 2030
• Starcloud submitted an FCC proposal for up to 88,000 satellites in 2026
• Adaspace launched 12 AI-Cloud satellites in May 2025
• Beijing institute’s constellation targets 16 gigawatts of power
• SpaceX plans to use $25 billion from an IPO for orbital AI projects

US-based Starcloud launched its first satellite equipped with an Nvidia H100 GPU in November 2025, marking the initial deployment of data-center-class GPU computing in orbit. The company conducted in-orbit training of a large language model the following month and later submitted a proposal to the FCC for a constellation of up to 88,000 satellites dedicated to orbital data centers.

Other technology companies are also pursuing similar projects. Alphabet’s Google announced the “Sun Catcher” initiative in November 2025, which involves building a space-based AI data center using solar-powered satellites and Google TPUs, with test launches planned for 2027. Nvidia-backed Starcloud aims to develop a space data center covering several square kilometers and targeting five gigawatts of power within five years. Orbit AI’s DeStarlink Genesis-1 demonstration satellite, launched in December 2025, carries NVIDIA AI processing hardware powered by space-grade solar panels and has been operating in orbit to validate the architecture for an Orbital Cloud constellation.

Despite these developments, technical and logistical barriers remain. China has not yet achieved a fully successful reusable rocket program, which limits its ability to launch and maintain space-based AI infrastructure. Industry leaders have also pointed to challenges such as temperature fluctuations, radiation exposure, and the need for effective radiative cooling systems.

Industry reaction

Nvidia CEO Jensen Huang stated that space-based AI computing is currently a “dream,” referencing the technical and cost challenges involved, including the need for advanced cooling and protection from space conditions.

OpenAI CEO Sam Altman said that space-based data centers are unlikely to be practical at scale this decade, citing high launch costs and maintenance difficulties.

As multiple organizations continue to invest in research and pilot missions, the competition to establish operational AI data centers in orbit is ongoing, with both public and private actors seeking to overcome current obstacles and expand computing capabilities beyond Earth.

* This article is based on publicly available information at the time of writing.