On May 14, the China Aerospace Science and Technology Corporation confirmed the successful launch of a dozen specialized satellites into orbit. But these machines are not beaming down internet connections or snapping photos for military surveillance. Built by the state-backed firm GuoXing Aerospace, they represent the first functional nodes of the Star Computing program, an ambitious plan to construct an artificial intelligence supercomputer entirely in space.
744 Trillion Operations Per Second in a Vacuum
Each of the 12 orbiters launched this week by GuoXing Aerospace possesses a computing power of 744 trillion operations per second (TOPS). That is serious hardware normally reserved for temperature-controlled server farms, now floating in the freezing vacuum above Earth. Instead of sending raw data back down to ground stations for processing, this network will do the heavy lifting right where it sits.
The satellites communicate using laser inter-satellite links capable of 100 Gbps speeds. They talk to each other directly, sharing the workload of complex AI tasks without needing a constant tether to Chinese soil. This marks a fundamental shift from traditional satellite operations, which usually act as simple messengers beaming information to heavy-duty computers on Earth.
By analyzing data directly at the source in orbit, the constellation filters out petabytes of useless information before it ever hits the downlink. If a surveillance satellite takes thousands of photos of an empty ocean, the orbital AI can simply discard them and only transmit the images containing specific naval vessels. This drastically reduces the bandwidth required and speeds up real-time decision making for operators on the ground.
Moving actual computation to low Earth orbit offers several technical advantages over traditional setups:
- Reduces reliance on overburdened terrestrial power grids and local utilities
- Eliminates the need for millions of gallons of cooling water
- Filters petabytes of raw imagery before choking limited downlinks
- Operates independently using unrestricted solar power arrays
Escaping the Terrestrial Data Center Bottleneck
Cooling the heat generated by artificial intelligence is exceedingly expensive. Ground-based AI growth in China currently faces severe bottlenecks in power supply and water access for server farms. Training and running large language models requires vast amounts of electricity, which strains local grids and creates a substantial carbon footprint.
By putting the processors in space, engineers can use the harsh environment to their distinct advantage. Orbital edge computing nodes run on continuous solar power gathered without atmospheric interference. More importantly, they radiate their excess heat directly out into the cold expanse of space, bypassing the need for complex liquid cooling infrastructure entirely.
“It is the first substantial flight test of the gambit.” – Jonathan McDowell, Astrophysicist at the Harvard-Smithsonian Center for Astrophysics.
The state-backed developers at GuoXing Aerospace have mapped out a long timeline for this ambitious project. Their ultimate goal is to build a network of 2,800 satellites by 2035, consisting of 2,400 inference nodes and 400 training nodes. If completed, the Star Computing network would reach a target capacity of 100,000 petaflops, creating the largest decentralized computer ever built by humanity.
The Strategic Gap Between Internet and AI
SpaceX commands the sky when it comes to raw numbers, boasting over 6,750 Starlink units orbiting the globe as of early 2025. But comparing Starlink to China’s new initiative is missing the fundamental point. American mega-constellations are designed primarily to bounce broadband internet to rural homes, maritime vessels, and commercial flights.
Even China’s other major space project, the Thousand Sails constellation, follows this traditional commercial model. The first 18 satellites of that broadband network launched in August 2024, aiming to eventually put 15,000 internet routers in orbit to compete directly with Elon Musk’s enterprise.
The Star Computing network serves a completely different, purely strategic purpose. It operates as a national asset meant to run machine learning models, analyze battlefield simulations, forecast climate events, and process real-time surveillance without relying on vulnerable ground infrastructure.
| Constellation Program | Primary Network Purpose | Current Orbiters (Early 2025) | Final Target Size |
|---|---|---|---|
| SpaceX Starlink (United States) | Broadband Internet Access | 6,750+ | 30,000+ |
| Thousand Sails / Qianfan (China) | Broadband Internet Access | 18 | 15,000 |
| Star Computing (China) | Space-Based AI Processing | 12 | 2,800 |
Economic Push Behind the Strategic Emerging Sector
Building a digital brain in the sky requires an astronomical amount of capital. Fortunately for the engineers at the China Aerospace Science and Technology Corporation, the financial floodgates are wide open. The National People’s Congress recently classified commercial space as a strategic emerging sector, signaling high-level state support and guaranteed funding for private aerospace projects.
Local municipalities are throwing significant weight behind these orbital ambitions. The Shanghai municipal government recently directed a 6.7 billion yuan investment into local satellite clusters to build the necessary manufacturing infrastructure. This influx of cash allows companies to iterate quickly, turning theoretical research into flight-ready hardware in a matter of months rather than years.
Back in November 2014, China’s State Council published a sweeping policy known as Document 60, which opened the strictly guarded space sector to private investment for the first time. Today, the landscape is crowded with ambitious start-ups and state-funded tech clusters working in tandem to secure orbital real estate and valuable radio frequencies before international competitors can claim them.
Harder to Jam, Harder to Destroy
During a major conflict, terrestrial communication cables and centralized data centers become immediate prime targets for cyberattacks and missile strikes. If an adversary destroys the cooling infrastructure of a ground-based supercomputer, the entire system melts down in minutes. Physical infrastructure is a major vulnerability in modern warfare.
A decentralized brain floating in orbit is much harder to take down. Space-based AI can run beyond the immediate reach of conventional weapons, maintaining critical operations even if Earth-based systems are completely cut off. Because the satellites communicate via laser links, they are also deeply resistant to traditional radio frequency jamming techniques used to disrupt military communications.
The United States military is well aware of this strategic shift. Just last month, an American satellite was tracked flying unusually close to Chinese orbiters in low Earth orbit. That maneuver was not a navigational error; it was deliberate surveillance and a clear message that the U.S. Space Force is actively monitoring these technological developments.
To maintain operational security, the Star Computing constellation relies on several built-in redundancies:
- Hundreds of backup nodes to replace damaged units instantly
- Direct laser communication that bypasses vulnerable ground relays
- Autonomous decision making capabilities independent of human operators
- Distributed computing that prevents a single point of network failure
GuoXing Aerospace has yet to reveal exactly how frequently the remaining 2,788 satellites will be deployed. However, the nation does not typically drag its feet on technology rollouts of this magnitude, especially with global competition watching their every move. The U.S. and China are actively competing step-for-step in both advanced processing and orbital assets, pouring billions into securing the high ground.
The overlap between advanced machine learning and orbital mechanics is no longer just a theoretical concept discussed in academic research papers. The hardware is actively operating above our heads right now, silently crunching data and proving that the next great technological leap will not happen on the ground. As other nations scramble to match this capability, the true scale of the #SpaceRace is expanding from simple communication networks to a battle for #ArtificialIntelligence dominance in the stars.
