In the contemporary era, space computing has emerged as a pivotal technological frontier, actively pursued by numerous nations. China’s Three-Body Computing Constellation stands as the world’s inaugural space-based computing network achieving full-orbit interconnection. With a total computational capacity of 5P OPS (50 quadrillion operations per second), the constellation has already deployed multiple large-scale AI models.

At the recent 2026 National Science and Technology Workers' Day Zhejiang Main Event, Li Chao, Director of the Space-Based Computing Systems Research Center at Zhejiang Lab and Chief Technologist of the Three-Body Computing Constellation, shared insights into the constellation’s development journey.

After nine months of intensive research and development, the lab-led “Three-Body Computing Constellation” project successfully completed crucial in-orbit validations. On May 14, 2025, a Long March 2D carrier rocket propelled 12 AI-powered intelligent computing satellites into their designated orbits, marking the official birth of the world’s first space computing constellation with full-orbit interconnection capabilities.

This milestone signifies China’s formal transition from conceptual verification to operational deployment in space computing. As of now, the constellation boasts a total computing power of 5P OPS (50 quadrillion operations per second), with a maximum single-satellite computing power of 744TOPS, ranking it first globally in terms of in-orbit computing capacity.

Traditionally, space data processing has adhered to a “space-based sensing, ground-based computing” model. Satellites capture vast amounts of imagery in space; however, due to limitations in satellite-to-ground transmission bandwidth and the geographical distribution of ground stations, less than 10% of this data can be transmitted back to Earth. The remaining 90% is often discarded due to the inability to transmit it promptly.

For ultra-high-resolution remote sensing satellites, the hundreds of GB of raw data captured in a single pass cannot be fully transmitted during the brief 10-minute window when they are in range of ground stations. This inherent time lag renders satellites ineffective in time-sensitive scenarios, such as emergency disaster response and real-time monitoring.

The Three-Body Computing Constellation aims to disrupt this status quo by shifting data processing from the ground to space, enabling “space-based data processing in space.”

To achieve this transformative leap, the technical team focused on breakthroughs across three key dimensions. Firstly, they enhanced in-space computing power by equipping satellites with domestically produced aerospace-grade chips and onboard intelligent computing servers, enabling real-time data processing in orbit. Secondly, they established satellite interconnection, allowing satellites to share data and computing resources akin to a local area network through inter-satellite laser communication technology, thereby reducing reliance on ground-based remote control. Thirdly, they deployed large-scale AI models, such as Alibaba’s Qwen3, on in-orbit satellites, achieving rapid in-orbit updates within two days. This evolution transforms satellites from mere data collectors into intelligent nodes capable of autonomous analysis and decision-making.

In the global race for space computing supremacy, China’s manufacturing supply chain has demonstrated remarkable resilience. Unlike SpaceX’s highly vertically integrated and closed system in the United States, China’s commercial aerospace sector thrives on a distributed and open industrial chain collaboration. For instance, in this mission, Zhejiang Lab collaborated with multiple institutions, including Guoxing Aerospace, to swiftly complete the entire process from design to launch, leveraging mature industrial support. As a core enterprise in the industrial chain, Guoxing Aerospace has revolutionized satellite development by shortening the entire cycle by 80% through an “AI design + additive manufacturing” model, leading to explosive growth in its satellite platform business in 2025. This cost-effective and agile manufacturing capability forms the bedrock supporting the deployment of a constellation comprising thousands of satellites.

The current deployment of 12 satellites is merely the beginning. According to the plan, the ultimate vision for the Three-Body Computing Constellation is to construct an in-orbit network composed of 1,000 satellites, with a total computing power nearing the 1E OPS (100 exaflops) milestone. This ambitious goal aims not only to meet domestic demand for low-latency, wide-coverage computing services but also to establish a globally accessible space artificial intelligence infrastructure. When computing power becomes as ubiquitous as utilities in space, future applications such as autonomous driving, deep space exploration, and planetary science research will heavily rely on this omnipresent “space network.” In this transformative phase of reshaping space exploration paradigms, China has secured a pivotal ecological niche by leveraging its first-mover advantage.

“Although we pioneered this endeavor, we cannot claim to be the best, as space computing is still in its nascent stages,” remarked Li Chao. He reiterated that the ultimate objective of the Three-Body Computing Constellation is to construct a space computing infrastructure comprising 1,000 satellites in orbit, with a computing power of 1E OPS (100 exaflops).

“We aspire to build a globally open space AI infrastructure and revolutionize the paradigm of space scientific research through computing and intelligence,” emphasized Li Chao. He underscored that the journey to Mars would be unfeasible without the indispensable companionship of computing and AI technologies.