A few days ago, CGTN ran a piece titled “Race to orbit: Why ‘computing in space’ is the world’s next tech battleground.” The headline says it all. But since that article dropped, the race has only gotten hotter. Here’s what’s actually happening right now — and why it matters.
The problem Earth can’t solve anymore
AI is eating compute power at an alarming rate. Data centers already gulped down about 1.5% of global electricity in 2024, and that number is climbing fast as AI deployment accelerates. Cooling those chips? Even harder. Record heat waves aren’t helping.
So where do you put compute when Earth is running out of cheap power and cool air? Orbit.
What space computing actually means
Forget the old model where satellites act like “cameras that only take pictures” — collect raw data, beam it all down, wait hours or days for ground supercomputers to process it. Industry estimates suggest less than 10% of collected data ever gets used that way.
Space computing flips this. Satellites get radiation-hardened chips, onboard servers, and storage. They process data in orbit and send back only the valuable results. Bandwidth consumption drops by over 90%. Response times shrink from hours to seconds.
Who’s going for it — and how big
SpaceX kicked things off in a big way. On January 30, the company filed with the FCC to deploy up to one million satellites as orbital data centers. Elon Musk’s vision? Solve AI’s terrestrial bottlenecks by putting servers in space, where solar power runs nearly 24/7 and the vacuum of space handles cooling for free.
China didn’t wait. Just weeks later, Beijing vowed to develop its own gigawatt-class space digital-intelligence infrastructure. And they’re already delivering.
In May 2025, China launched the first 12 computing satellites of the “Three-Body Computing Constellation.” By February 2026, the constellation had deployed 10 AI models in orbit and achieved inter-satellite networking among six satellites. One of those models — an 8-billion-parameter remote sensing AI — conducted an infrastructure census across 189 square kilometers in northwest China, automatically identifying stadiums and bridges even under heavy snow.
The lab behind it says once 1,000-plus satellites are up, the constellation will hit 100 quintillion operations per second.
GuoXing Aerospace is pushing even further. The company’s “Star-Compute” plan aims for 2,800 computing satellites by 2035 — 2,400 for inference, 400 for training. Their second and third satellite clusters are scheduled for deployment this year, with a 1,000-satellite network targeted by 2030.
The infrastructure catch-up
On June 29, 2026, Beijing unveiled the Beijing Space Computing Innovation Center — China’s first platform dedicated to linking the entire space computing chain. It covers six priority areas: chip design, high-performance payloads, satellite platforms, large model deployment, integrated space-ground cloud tech, and application innovation. An innovation alliance with 108 founding members — universities, research institutes, state-owned enterprises, and private firms — was established alongside it.
So what’s the takeaway?
This isn’t sci-fi. The first 12 satellites are already in orbit, processing real data. AI models are running in space right now. Companies are racing to deploy thousands more. The question isn’t whether computing moves to orbit — it’s who gets there first and at what scale.
Space offers unlimited solar energy, free cooling, and global coverage. Ground data centers offer… rising electricity bills and water shortages. The math writes itself.
The next few years will determine who controls the computational backbone above our atmosphere. And unlike the first space race, this one isn’t about flags and footprints. It’s about who powers the AI revolution.