In 2026, space computing emerged as an important new frontier in commercial space. What problems is it intended to solve, and where is the industry heading? Hello Space has launched its Space Computing Interview Series, speaking with leading Chinese players to examine the real progress being made in this emerging sector.
Our guest in this interview is Xie Hongjun, chairman and general manager of i-Space’s Shanghai-based space computing subsidiary.
Key Quotes
- Space computing is not a single-choice question. It is a multiple-choice question—and perhaps even a mandatory one.
- A rocket company’s real advantage lies in its ability to design rockets, satellites and computing nodes together from the perspectives of launch capacity and space systems engineering.
- Real validation is not about proving that computing power can be sent into space. It is about proving that computing power in space can become a product.
- The real business model for space computing may not be selling weather data, but selling computing power, timeliness and data-processing capabilities.
- Space computing is not something that one company can build alone. It requires an entirely new industrial ecosystem.
- We plan to spend five years exploring space computing, ten years developing it on a small scale and twenty years pursuing development on a much larger scale.
- Space computing requires genuine, in-orbit demonstration and validation at a meaningful scale.
Since private capital was formally allowed to enter China’s space sector in 2015, companies specializing in rockets, satellites, communications, navigation and remote sensing have emerged one after another. In the industry’s early years, most of these companies focused on their own specific market segments.
By 2026, however, the rapid growth of artificial intelligence had turned space computing into a major new focus. A growing number of companies began entering the field—not only satellite manufacturers and computing-payload developers, but rocket companies as well.
In the first half of this year, i-Space established a subsidiary in Shanghai, positioning space computing as an important growth engine beyond its rocket business. Xie said during the interview that the computing demand created by AI may be pushing commercial space beyond “sending things into space” and toward a new stage of building infrastructure in space.
In this interview, we sought to understand space computing from the perspective of a rocket company. In Xie’s view, space computing cannot simply be understood as adding an AI chip to a satellite. It is closer to an entirely new form of infrastructure.
Within this system, chips provide computing power, satellites serve as computing nodes, laser communications connect those nodes, power and thermal-control systems keep them operating, and rockets deliver the infrastructure into orbit. Models, data and applications are ultimately required to turn the system into something useful.
This means that once space computing enters the industrialization stage, its participants will extend far beyond aerospace companies. The semiconductor, AI, cloud computing, energy and communications industries will all become involved. This is also why i-Space’s space computing subsidiary believes the field is worth pursuing for five years, ten years or even longer.
Traditionally, rocket companies may appear to have little connection with space computing. Xie’s reasoning, however, is precisely the opposite: if future orbital data centers require large numbers of satellites, then launch capacity will become one of the industry’s most fundamental resources.
Building data centers in space requires rockets capable of providing low-cost, large-scale and reliable launch capacity. Otherwise, no matter how many computing satellites are manufactured on the ground, they cannot be placed in orbit. Launch capacity will therefore directly determine whether space computing can ultimately become a viable business.
Xie is also well aware that solving the launch problem does not solve every challenge. He repeatedly emphasized that space computing is a systems engineering undertaking spanning aerospace, semiconductors, distributed computing, AI, communications, energy and other fields.
A rocket company’s advantage is that, as the provider of the vehicle that ultimately carries computing infrastructure into orbit, it can naturally consider rockets, satellites and computing infrastructure within a single engineering system.
Although space computing has become a popular subject and many companies are discussing validation missions, computing satellites and constellations, Xie believes that the industry as a whole remains at an early stage. Engineering validation is far more difficult than it may appear.
He proposed a specific benchmark for validating space computing: a satellite must successfully run through the complete chain of power supply, chips, heat dissipation, distributed computing, models, data and applications before it can be considered to have passed the validation stage.
Ultimately, space computing will be judged by whether it can deliver genuinely usable computing power—and whether customers are willing to pay for it. At the industry’s current stage, a long road of technological breakthroughs and engineering validation still lies ahead before large-scale commercialization can become a reality.
In this discussion, we therefore focused on several questions: Why would a rocket company enter space computing? What advantages does a rocket company have over a satellite manufacturer? What constitutes genuine in-orbit validation of space computing? Who would be served by computing constellations comprising hundreds or even more satellites? And in the longer term, will space computing resemble national infrastructure, or can it become a viable business for commercial companies?
The following is an edited transcript of Hello Space’s interview with Xie Hongjun, chairman and general manager of i-Space’s Shanghai-based space computing subsidiary. Portions have been organized and edited for clarity.
Why Is a Rocket Company Moving Into Space Computing?
Hello Space: Why has a rocket company suddenly decided to enter space computing?
Xie Hongjun: This was not a sudden idea. We were considering where our second growth engine—a new growth trajectory beyond the rocket business—might come from.
Several considerations led us to computing.
First, we believe this will be a highly integrated industry combining space systems engineering, the semiconductor industry and large AI models. Its applications may be closely connected with AI, while the form in which it can currently be commercialized is computing services—or, in other words, token generation.
In China, many people are reluctant to discuss or even oppose the idea of generating tokens in space. We believe, however, that this will form part of the infrastructure of the AI era. We therefore see it as something with tremendous potential—something worthy of long-term investment. That was our first consideration.
Second, rocket companies need a sustained and continuing flow of orders. Before you establish stable, sustainable and airline-like launch-delivery capabilities, where will those orders come from?
This is fundamentally an interdependent problem. Without orders, a product cannot iterate and mature. Yet the market will not buy an immature product. The logic at this second level is equally clear.
Hello Space: Is access to low-cost launch capacity the greatest advantage a rocket company has in space computing?
Xie Hongjun: From a cost perspective, the factor constraining the speed and scale of space computing is the cost of transportation beyond Earth. This is also an important cost difference between space computing and terrestrial data centers.
That does not mean rocket companies are necessarily better suited to the business. Space computing is an extraordinarily complex undertaking. It cannot be solved independently by people specializing only in space systems engineering, AI or conventional terrestrial data centers.
It requires a multidisciplinary organization led by people with highly integrated expertise.
Hello Space: In previous remarks, you estimated that space computing currently costs around five times as much as terrestrial computing. Why not wait until reusable rockets mature before entering the field? How did you choose this particular moment?
Xie Hongjun: Many things do not depend on having every condition perfectly in place. They depend on demand.
We have seen the direction in which AI is developing, and we have seen NVIDIA CEO Jensen Huang’s five-layer model for artificial intelligence. Many of the associated business scenarios can be implemented in space, so why should we not pursue them?
The data is in space, so the models will follow it into space, and computing power will follow as well. If there is not enough computing power, it can be produced in space. Why should we not build it there?
The key question is not whether the commercial space industry is completely ready. The demand already exists objectively. If the demand is there, we should act. Through that process, demand can also pull forward development on the supply side. I believe the two reinforce one another.
Hello Space: For commercial space companies, is space computing the industry’s greatest commercial opportunity? Could it even become larger than satellite internet and remote sensing?
Xie Hongjun: I would not venture to make a definitive judgment about the industry as a whole. Personally, however, I believe space computing is a long-term undertaking.
The largest area of AI investment today remains infrastructure. The imbalance between demand for tokens and the supply needed to generate them will continue to grow. Can terrestrial infrastructure be built quickly enough to meet that demand?
From this perspective, space computing is an unavoidable trend. I would not make a hasty prediction about its eventual scale, but I believe it represents an extremely large business opportunity and is worth pursuing.
What Advantages Does a Rocket Company Have in Space Computing?
Hello Space: From a commercial perspective, aside from owning rockets, what is a rocket company’s greatest advantage over a conventional satellite manufacturer when developing space computing?
Xie Hongjun: Satellite design must be coupled with rocket design through integrated launch vehicle–spacecraft engineering. This provides advantages in areas such as environmental requirements and the efficient use of available space.
The size and configuration of satellites are constrained not only by power, thermal management and chips, but also by engineering considerations: how many satellites can a rocket launch at once, and what satellite dimensions are most practical?
That is an important distinction between rocket companies and other types of companies.
Second, returning to launch capacity, if the development of our rockets eventually gives us a certain level of transportation capability between Earth and space, launches of our own satellites will further support the expansion of rocket production while improving stability and reliability.
The two businesses reinforce one another. Internal coordination is also easier because they are our own rockets and our own satellites.
What Are the Real Bottlenecks in Space Computing?
Hello Space: At present, what do you see as the largest barrier to building space-based data centers? Is it AI satellites, chips or launch capacity?
Xie Hongjun: None of these challenges is simple.
China’s planned low-Earth-orbit communications constellations involve tens of thousands of satellites. Planned constellations outside China are even larger, reaching into the hundreds of thousands. I believe computing infrastructure will also require a substantial number of satellites.
Academician Yang Hong has said that a single satellite may need more than 100 kilowatts of power. Only when a satellite reaches a certain computing scale will its computing capabilities become genuinely effective and usable.
If we can place highly concentrated computing capacity on this scale into orbit, it may reduce the complexity of distributed computing architectures and designs, as well as the scheduling architecture of future computing systems.
In terrestrial computing centers, different computing units are effectively connected to one another, providing extremely high transmission rates and very low latency. Satellites, however, may be separated by hundreds or thousands of kilometers. At present, inter-satellite links appear to depend on laser communications.
China has so far validated space-based inter-satellite links at approximately the 100-gigabit level. This is where the differences between orbital and terrestrial infrastructure begin to emerge.
Another difference is dynamism. Unlike terrestrial computing centers, which have stable network topologies, network topology in space is highly dynamic.
Simply getting the system into space is already a major undertaking. The industry should therefore avoid making sweeping claims about supposed breakthroughs. We need greater patience and a stronger sense of responsibility, proceeding one step at a time.
What Constitutes Genuine Validation of Space Computing?
Hello Space: Many companies currently say they are conducting validation. You have argued that validation must be real engineering validation at a meaningful scale. What specific criteria would you use to evaluate it?
Xie Hongjun: First, there is the prerequisite of power. Second, what kind of AI chips will be used to build the hardware system, and can those chips dissipate heat safely and reliably under different operating modes?
Third, what communications infrastructure will be used to operate the entire system, and what data will it process? Fourth, are there customers willing to pay for the capabilities or services created by space computing?
These questions determine how large the satellite must be, how large its solar arrays must be, how much power it can generate, how its thermal-management system will work, what chip architecture will support its distributed computing system, and what products built from what kinds of models can serve its users or platforms.
Hello Space: In terms of power, how much do you believe computing satellites will need over the next three to five years?
Xie Hongjun: Ten kilowatts may be the minimum requirement today. Academician Yang Hong has proposed more than 100 kilowatts, but that would already require a very large satellite.
Launch capacity remains a constraint today. In the medium to long term, satellites can become larger. The greater the centralized computing capacity aboard a single satellite, the broader the types and scale of computing tasks it can process.
Given current rocket capabilities, the available area for power generation, the choice of chips and the selection of solar cells are all critical.
We have discussed reducing launch costs through reusability, but satellite costs must also come down. When those costs are broken down, the satellite platform and chips are the main components.
Chip costs will need to be reduced through scale in the semiconductor industry. A large portion of satellite cost comes from the power system because this is an energy-intensive business. I believe the first step should be validation at approximately the 10-to-20-kilowatt level.
What Is i-Space’s Space Computing Subsidiary Planning to Do?
Hello Space: You have described space-based computing centers as a capital-intensive business that may ultimately be led by state-backed organizations. How do you assess that possibility? What does i-Space’s space computing subsidiary plan to do—and what does it not plan to do?
Xie Hongjun: Our first priority is to build the rockets and satellites well.
As for chips, whether we eventually build our own computing infrastructure or purchase chips to assemble it—and who we will sell the resulting services to and how we will sell them—are all questions open for discussion.
As a commercial and privately owned enterprise, we understand our advantages and the resources we can mobilize. Our priority is to do the work immediately in front of us well.
Hello Space: Aside from rockets, what has the subsidiary already begun developing?
Xie Hongjun: Some time ago, we worked with a Chinese AI chip company to present a space computing use case.
In the future, we may also discuss with the wider industry how to pursue the areas that we consider important. Simply building the satellites properly will already require a tremendous amount of effort. Clearly explaining the entire system also requires substantial professional depth.
As for infrastructure development, we cannot do everything ourselves. We will begin by approaching the problem from the perspective we know best: space systems engineering.
Hello Space: The company was only recently established and remains in the system-architecture and concept-development stage. Have you nevertheless set specific targets—for example, what you hope to accomplish within three or five years?
Xie Hongjun: We have not made them public, but we do have plans.
With the small-scale experiments we are currently capable of conducting, we may launch several satellites to validate key technologies of shared interest across the industry.
If the validation proceeds successfully, we will establish a small-scale network. Once that network has been completed, begins generating revenue and perhaps even reaches break-even, we will move on to large-scale constellation deployment.
Hello Space: How large would that initial network need to be? You mentioned that it might become profitable at the small-network stage. Who would its primary customers be, and how could it achieve profitability?
Xie Hongjun: Even this small-scale network would need at least several hundred satellites.
Once it reaches that size, it could support a wide range of service scenarios. In addition to space-based AI, it could address other needs involving data timeliness, accessibility and broader coverage.
These are all areas that we would explore through the initial network. We also believe that the concept of space-based token factories deserves deeper consideration across the industry.
What Is the Business Model for Space Computing?
Hello Space: Beginning in the second half of this year, computing capabilities are expected to be added to every newly launched satellite. We believe computing will eventually become standard equipment on every satellite. If that happens, will there still be a market for small computing constellations and dedicated space computing networks?
Xie Hongjun: Adding computing capabilities to every satellite is fundamentally different from creating a network that can provide computing services through a constellation.
I also do not believe it is realistic to equip every satellite with four or five laser communications terminals.
If we are talking about multiple satellites or a constellation, should every satellite become a fully integrated spacecraft with remote-sensing, computing and communications capabilities? I do not believe every satellite can be designed that way.
Communications infrastructure and computing infrastructure will eventually become distinct forms of infrastructure, with different participants performing their respective roles.
Adding computing power to satellites today can substantially improve application timeliness and expand service reach. It is an interim solution. To solve the problem comprehensively, however, I believe much more powerful infrastructure will ultimately be required.
What Is the Industry’s Most Common Mistake?
Hello Space: What do you believe is the biggest misconception about space computing today?
Xie Hongjun: The greatest misconception may be the doubt surrounding the entire idea.
At the beginning of any new field, many different opinions emerge. Truth can only be established through the exchange and debate of different viewpoints.
China’s current chip capabilities will certainly constrain the speed at which this infrastructure can develop. Nevertheless, I believe we need to build momentum so that people recognize that this is worth pursuing—and worth pursuing over the long term.
Hello Space: We have also interviewed the leaders of several other companies in the industry. Many believe that the greatest risk may be that the sector has become overheated and that the entire industry is increasingly impatient, affecting both talent and technology strategies. Do you see this enthusiasm as beneficial or harmful?
Xie Hongjun: First, enthusiasm is good for the industry. It is positive when the primary market is willing to invest in the sector. If commercial space companies enter the public markets, this enthusiasm may also support the development of the secondary market.
The downside is that enthusiasm can cloud people’s judgment. Companies may assume that market attention means they will automatically perform well—or even dominate the industry.
I believe we need many different approaches and forms of exploration. Companies should not become complacent. They need innovation and clear targets—for example, specific cost goals for rockets and satellites.
Anyone can promote a concept. You can quickly launch a satellite, or work with a partner to place a computing payload into orbit, and then claim that you have completed the entire process.
But can you deploy computing capacity at a meaningful scale, carry useful general-purpose or specialized models into orbit, connect them immediately to operational business systems and deliver a product that customers are willing to pay for? Can you achieve that through a small-scale validation program?
I believe that is extremely difficult. Validation must still take place at a meaningful scale.
We must pursue what the space industry regards as genuine, in-orbit demonstration and engineering validation at a certain scale—not superficial validation that merely scratches the surface.
That requires a thorough understanding of the industry and adequate financial support before meaningful progress can truly be made.
Perspective
The commercial space narrative is shifting from “sending things into space” to “building infrastructure in space.”
The establishment of i-Space’s space computing subsidiary represents a new line of reasoning: a rocket company is attempting to use launch capacity as leverage to enter the long-cycle space computing sector. In this interview, Xie discussed possible validation standards for space computing and acknowledged that realizing the vision will require participation from multiple industries.
Whether space computing ultimately becomes a major commercial business or more closely resembles national infrastructure may not become clear for another five years—or even longer. But at the very least, some companies have begun addressing the question through the logic of engineering.
