The race to deploy AI infrastructure in low-Earth orbit is intensifying, with China currently leading the development of “space supercomputers.” Supercomputers designed to run AI programs require massive data centers that consume enormous amounts of energy—and even more to keep them cool.

Space-based supercomputers, however, could harness solar power and the natural cold of space for passive cooling, reducing both energy use and operational costs.

Chinese organizations, including the Institute of Computing Technology, Zhejiang Lab, and Guoxing Aerospace, have already launched satellite constellations capable of computing, setting the stage for a full-fledged orbiting supercomputer.

At the same time, U.S. tech giants like Elon Musk, Jeff Bezos, and Google CEO Sundar Pichai are investing heavily in competing projects, ranging from AI-enabled Starlink satellites to experimental satellite-based data centres.

A potential front-runner is the Nvidia-backed startup Starcloud, which recently launched its Starcloud-1 satellite carrying an H100 GPU.

This milestone marked the first AI model, NanoGPT, trained in space, and the first time an Nvidia GPU ran a large language model in orbit.

Advocates say space-based AI data centres could dramatically reduce energy and water use while lowering greenhouse gas emissions by leveraging solar power.

Despite significant technical hurdles—such as rocket vibrations, radiation, microgravity, and extreme temperatures—experts predict operational orbital supercomputers could be realized by the 2030s.

The pressing question remains: who will be the first to achieve sustainable, large-scale AI computing in space?