May 30, 2025

China Is Building The First AI Supercomputer in Space

China Is Building The First AI Supercomputer In Space
A Long March 2D rocket launch recently began the Chinese quest to host the world’s first AI supercomputers from space. This picture is a March 2023 launch of Hongtu-1 satellites (Credit: WikiMedia Commons).

Just after midnight on May 14, a Long March 2D rocket carried 12 identical satellites into low Earth orbit. These 12 craft form the first “node” of the proposed 2,800-satellite fleet called the Three-Body Computing Constellation, an ambitious project led by Chinese start-up ADA Space and the government-affiliated Zhejiang Lab. The full plan calls for all satellites to be deployed over the next several years. If everything stays on schedule, the fleet will become the world’s first large-scale computer network that lives entirely in space.

Why build computers in orbit?

Running advanced chips takes energy and produces heat. Data centers on Earth solve the problem with huge air-conditioning plants and, quite often, enormous volumes of fresh water. Orbit offers a different set of tools. Solar panels provide a steady power source almost all the time (and with much greater efficiency than on the ground where sunlight is first filtered by the atmosphere), and the vacuum of space lets heat radiate away without fans or chilled pipes. By tapping those natural advantages, ADA Space says an orbital data center should consume far less Earth-based electricity than a similarly capable facility on the ground.

That environmental angle matters because artificial intelligence models keep growing. Former Google chief executive Eric Schmidt told U.S. lawmakers in April that new data centers might need as much extra power this decade as a country the size of Japan consumes today. Finding clever ways to keep computers cool—whether in cold northern climates, under the ocean, or up in space—is fast becoming part of the global energy conversation.

“It’s a good time to think about how we can put AI into space, not just in your laptop or cellphone,” Wang Jian, director of Zhejiang Lab, said at the Beyond Expo tech conference in Macau in the South China Morning Post (SCMP) reported. “Space has, again, become the frontier for us to think about what we can do in the next 10, 20, or 50 years.”

Computing power matters

Each satellite launched carries an AI model with eight billion adjustable “parameters,” roughly speaking, the dials that let a neural network learn patterns. For comparison, the language model inside a modern smartphone assistant usually sits below one billion. ADA Space says one of its orbiting computers can perform about 744 trillion operations per second. Connect a dozen of them, and the figure rises to five quadrillion. If the whole 2,800-satellite constellation comes online as planned, the ultimate capacity could rival or surpass the fastest supercomputers housed in sprawling buildings on Earth.

Most satellites in the Three-Body network will point their sensors toward Earth, serving civil, commercial, and possibly military users. However, one of the initial 12 carries an X-ray instrument to spot short-lived flashes from distant gamma-ray bursts. Catching those events quickly helps astronomers swing other telescopes toward the source. The payload highlights another selling point of space-based computing: when a phenomenon lasts only seconds, every moment saved on data transfer can sharpen scientific results.

ADA Space, founded in 2018, specializes in artificial intelligence applications for satellites and has filed to list shares on the Hong Kong stock exchange. Zhejiang Lab opened a year earlier as a partnership among the provincial government, Zhejiang University, and e-commerce giant Alibaba. The pair hopes to speed up everything from chip design to satellite manufacturing by combining start-up agility with state resources.

The venture aligns with China’s broader push to expand its high-tech infrastructure. National policy documents describe AI, clean energy, and space services as key growth areas through 2030. Officials view on-orbit computing as a way to leapfrog ground-based network bottlenecks and stake out a leadership position in a still-nascent field.

<!– Tag ID: zmescience_300x250_InContent_3

[jeg_zmescience_ad_auto size=”__300x250″ id=”zmescience_300x250_InContent_3″]

–>

A new global space race arena

The idea of “cloud computing” above the clouds has captured attention elsewhere. U.S. firms Axiom Space and Jeff Bezos’s Blue Origin have floated plans for orbital data centers, though neither has yet launched dedicated hardware. By moving first with an operational cluster, China has set a benchmark others must now meet or exceed.

ADA Space executives say they intend to add new satellites every few months. Once roughly 300 are flying, the company will open a pilot service covering most of the globe between about 500 and 1,000 kilometers altitude. Software teams are already testing ways to split a task across dozens of satellites, pool the results, and deliver a finished product seconds after the first sensor reading.

Building and launching thousands of satellites is hard enough; steering them clear of orbital debris and coordinating radio channels adds another layer of complexity. Regulators worldwide are still writing rules for mega-constellations, and low-Earth orbit is growing busier by the week.

But even with that in mind, the launch puts China in the lead for orbital supercomputers in the newest space race.