China wants to build gigawatt-scale AI data centers in space

Terrestrial power grids are buckling under the weight of AI clusters, so space agencies and aerospace giants are taking the only logical next step: looking up.

Chinese state media reported that China plans to launch space-based AI data centers over the next five years. The goal, outlined by the country's primary space contractor, China Aerospace Science and Technology Corporation (CASC), is to build gigawatt-class space digital-intelligence infrastructure.

The announcement lands just days after Elon Musk pitched virtually the same vision, officially turning Earth orbit into the next battleground for artificial intelligence.

Moving servers into orbit

According to CASC, these orbital data centers will integrate cloud, edge, and terminal capabilities. The idea is to deeply fuse computing power, storage, and transmission bandwidth so data beamed from Earth can be crunched directly in space.

A December CASC policy document goes even further, outlining an industrial-scale "Space Cloud" powered by gigawatt-class solar hubs by 2030. The integration of space-based solar power with AI computing is set to become a core pillar of China's upcoming 15th Five Year Plan.

SpaceX is moving on an almost identical timeline. Musk revealed at Davos that SpaceX expects to tap funds from a planned $25 billion IPO this year to build orbital AI hubs.

Musk's math is straightforward: solar panels in orbit capture five times more energy than those on the ground. He called the move a no-brainer, predicting space will be the cheapest place to run AI within two to three years.

The reusable rocket bottleneck

Dreaming about data centers in the sky is one thing, but hauling tons of compute into orbit is entirely another.

China has serious momentum, logging a record 93 space launches last year. Just this week, the Chinese Academy of Sciences opened its first School of Interstellar Navigation to train engineers in deep space propulsion and navigation.

However, China's glaring bottleneck remains reusability. Unlike SpaceX, which dominates low Earth orbit thanks to the reusable Falcon 9, China has yet to complete a successful reusable rocket test.

Without cheap, repeatable launch capability, lofting heavy server clusters and orbital solar arrays remains an extraordinarily expensive proposition, regardless of state ambitions.

Why it matters

For everyday developers, space-based inference is obviously not something you can deploy to tomorrow. But for Earth observation, defense, and climate modeling, processing raw satellite imagery in orbit instead of downlinking petabytes to terrestrial stations would be transformative.

For the tech industry, this marks the official fusion of the AI boom and the new space race. Land-based data centers are running out of power and water, making the unlimited solar power of orbit deeply attractive.

That said, we think building orbital compute hubs glosses over terrifying engineering realities. Space radiation destroys silicon, vacuum cooling is notoriously difficult, and hardware maintenance is practically impossible.

It looks like orbital AI will remain a high-stakes playground reserved exclusively for players who master dirt-cheap reusable rocketry first.