Deploying AI data processing capabilities directly onto satellites, often referred to as “space-based computing,” has become a prominent topic in the current AI landscape. Leaders from the aerospace, satellite, semiconductor, and energy sectors are viewing this field as a promising new avenue for growth within the commercial space industry.
There are several compelling reasons behind the growing interest in space-based computing, according to experts in the field. The first is the limited availability of orbital slots, with only around 100,000 usable low-Earth orbit positions worldwide. While China controls hundreds of these slots, mega-constellations like Starlink operate over 10,000, highlighting the competitive rush to secure this valuable real estate early on. Early deployment can serve as a strategic reserve, offering competitive advantages for satellite operators and providing a new revenue stream for satellites that primarily serve remote sensing functions.
Second, space-based computing is designed to complement ground-based systems rather than replace them. It provides essential backup capabilities during natural disasters, network failures, or in remote locations such as deserts, oceans, mountains, and forests where ground stations cannot operate effectively.
The third key driver is energy. The high power consumption of data centers on Earth and the limited availability of renewable energy in certain regions, such as Shanghai, constrain data processing capacity. Space-based computing offers an alternative solution to this bottleneck by utilizing energy sources in space, where green electricity can be more feasibly harnessed.
These technological ambitions are set against a supportive policy and market backdrop in the country. Last December, a dedicated commercial space department was established, followed by the release of an action plan targeting high-quality development of the sector by next year.
Internationally, private companies like one of the leading aerospace firms have gone public, raising significant capital and achieving nearly 9 million users on their satellite-based internet service, which generated about $11.4 billion in revenue last year—a profitable venture for the company.
Shanghai, with its strong talent pool and established supply chain, is positioned to develop space-based computing significantly. The city has a full-fledged commercial space ecosystem, similar to Beijing, supported by its aerospace design institutes. However, the main challenge involves figuring out who will pay for these services and establishing a sustainable billing model. The traditional pay-per-use models used in ground cloud computing may not directly apply, given the complexity of various workloads and their durations.
Companies involved in satellite manufacturing are leveraging recent technological developments. For example, radar satellites with onboard processing capabilities can analyze terrain changes in real-time, issuing early warnings hours before conventional systems can, which is particularly useful in remote or infrastructure-limited regions.
To further develop this industry, experts suggest expanding existing subsidy programs and establishing international data governance standards to facilitate cross-border data sharing, critical for commercial satellite operations. The country already demonstrated the feasibility of this technology last November when a team successfully deployed a large language model onto a space-based constellation, completing an inference task in under two minutes. Plans are in place to expand this constellation to around 2,800 satellites.
Satellite design faces challenges such as heat dissipation, with current systems reaching ten times the heat density of traditional satellites, and radiation exposure, which can cause errors or damage to chips. To counteract radiation effects, systems need to incorporate radiation-hardened components, redundancy, and error correction software.
While U.S. companies pursue similar goals—with SpaceX integrating data processing modules into their upcoming Starlink satellites and Google testing prototypes with advanced chips—the high cost of reaching orbit remains a significant barrier. The recent successful controlled recovery of a reusable first stage by rockets in China marks a major milestone, bringing operational costs closer to a tipping point. Industry estimates suggest that achieving launch costs around $200 per kilogram will be critical for broad commercialization—anticipated between 2030 and 2035—compared to current costs of around $3,000 per kilogram with existing systems.
Traditional large communication satellites cost millions of dollars to manufacture, making fleets of tens of thousands of satellites financially unfeasible at current prices. To make space-based computing more accessible, the industry aims to reduce satellite costs to the tens of millions and eventually into the single-digit millions of dollars, with strategies borrowed from the electric vehicle sector, such as standardization of satellite subsystems.
Thermal management constitutes nearly 30% of satellite costs, significantly higher than the 10-15% seen in conventional satellites. Unlike traditional satellite functions like communications or navigation, space-based computing centers on selling processing power rather than connectivity. The main challenge remains determining whether customers are willing to pay continuously for onboard computing services, which will ultimately drive the success of this emerging industry.