AI's Final Frontier: The Race to Build Space-Based Supercomputers by 2025

Space is becoming the next frontier for artificial intelligence. By late 2025, major projects are already underway to move AI infrastructure beyond Earth. Companies and research labs are racing to deploy powerful computing systems in orbit, driven by soaring demand and the need for sustainable solutions. Chinese firms are leading the charge in space-based AI. Zhongke Tiansuan has operated a space computer aboard a Jilin-1 satellite for over 1,000 days. Their latest system, Aurora 5000, features a high-performance domestic GPU and is set for orbital testing soon. Meanwhile, Zhejiang Laboratory launched the first 12 satellites of its Three-Body Computing Constellation in May 2025, forming a 'mini computing constellation' designed for distributed AI tasks.

Experts predict 2025–2027 will mark a turning point for the industry. The strain on Earth’s data centres—from power consumption to water usage—has pushed companies to explore orbital alternatives. Space offers constant solar energy, natural cooling, and reduced data transfer delays. Google’s *Project Suncatcher* aims to capitalise on these advantages by launching solar-powered AI data centres in low Earth orbit, with test satellites planned by 2027. Long-term ambitions are even bolder. China has announced plans for a central space data centre in dawn-dusk orbit, targeting a power capacity of over 1 gigawatt by 2035. Globally, the AI supercomputer sector is projected to expand massively, reaching around 2 × 10²² 16-bit operations per second by 2030.

The shift toward space-based AI is gathering speed. With operational systems already in orbit and major projects on the horizon, the next few years will likely see rapid growth. These developments aim to meet surging computational demands while easing pressure on Earth’s resources.