China races to build AI computers in space as Nvidia chip controls tighten on earth

With Nvidia chips increasingly out of reach, China is pushing AI computing into orbit. New “space supercomputers” promise colossal power beyond Earth’s limits.

Several Chinese firms are venturing into space with a view to exploring orbital AI computing platforms, especially as earthbound limits such as land, water, and grid limitations can be circumvented.

Zhongke Tiansuan (Comospace), a company founded in 2024, is currently promoting a “space supercomputer” in low Earth orbit.

It subsequently plans to expand more broadly by building a gargantuan space center, orbiting 700 to 800 kilometers above Earth, with colossal power capabilities exceeding 1 gigawatt.

The first exhibitive satellite could lift off before the close of this year, but more realistically will happen in early 2026.

As lofty as these ambitions sound, the president of Beijing Astro-future Institute of Space Technology (BAIST), Zhang Shancong, is staying humble, to stay in tune with the vision:

Its scale is modest, but we are taking this first small step.

There are three stages of deployment for Comospace, with the ultimate endeavour envisaged for 2035, whereby an imposing data center will take to the skies.

The Zhejiang Laboratory in eastern China is also utilizing AI for a mini-constellation containing 12 satellites, called “Three-Body,” which will take on a staggering 100 quintillion operations per second.

Moving AI computing into orbit has major energy and cooling conveniences, namely passive cooling and solar panel scaling involved with space vacuums, acting as a giant, passive refrigerator.

Downstairs on Earth, meanwhile, ambitions would be constrained by Earth’s large, albeit fossil-dependent energy grid.

It’s not all-or-nothing, however, as the harnessing of power can be split between in space and on earth. “With a computing constellation, part of the data can be processed in space and delivered straight to users,” said Li Chao from Zhejiang Laboratory.

In the States, as American export controls increasingly restrict AI-use GPU access, in particular Nvidia chips (though this may change), China is looking for non-traditional power scaling strategies, as space-computing becomes a kind of escape vector.

This could trigger an international AI-in-space arms race with the US, India, and potentially Japan poised to respond.

Unlock more exclusive Cybernews content on YouTube.