The team behind BeiDou has launched a startup to build a space-based data center. Interstellar Origin has completed two consecutive rounds of financing totaling tens of millions of yuan | 36Kr Exclusive

Text | A Zhi

A new player has entered the space computing power track.

36Kr has learned that Star Origin Aerospace Technology (Shanghai) Co., Ltd. (hereinafter referred to as "Star Origin"), a provider of space computing power satellite system solutions, has recently successively completed two rounds of financing, the seed round and the angel round, with a total scale of tens of millions of yuan.

The seed round investor is JIUHE Venture Capital. The angel round was jointly led by existing shareholders JIUHE Venture Capital, Meihua Venture Capital, and the Ceyuan Fund of Shanghai Science and Technology Innovation Group under Shanghai State-owned Investment, with Shanghai Angel Association participating in the form of co-investment. Yiwei Capital served as the exclusive financial advisor.

Founded in 2025, Star Origin focuses its business on space computing power satellites. Its core team is jointly formed by key backbones of the BeiDou-3 program from the Chinese Academy of Sciences and experts from Harbin Institute of Technology, boasting national-level satellite engineering R&D experience for more than 10 medium and high-orbit BeiDou-3 satellites and over 30 low-orbit internet satellites. Founder Sun Xiaolei has over 15 years of aerospace engineering experience, having previously served as the chief satellite system designer and chief engineer at the Innovation Academy for Microsatellites of CAS and leading commercial aerospace companies.

BeiDou-3 is the first independently constructed satellite navigation constellation in Chinese aerospace history that achieved global networking. Its 30 satellites operate collaboratively across multiple high and medium orbital planes, taking a full 11 years from the official project approval in 2009 to the completion of global networking and official service launch in 2020.

The BeiDou-3 R&D team at CAS that Sun Xiaolei was part of at the time undertook the development of half of the satellites. "Unlike other research-focused institutes, the Innovation Academy for Microsatellites is dedicated to satellite engineering. Additionally, competition mechanisms were introduced into the project, pushing everyone to compete on technology and cost, which drove the formation of a complete new supply chain system."

The experience of developing BeiDou-3 represents the capability to deliver a national-level constellation on time with guaranteed quality using minimal resources under extremely demanding requirements. This is the team capability that Star Origin's early investors value most.

In 2023, Sun Xiaolei chose to leave his public institution position to serve as the chief satellite engineer at a commercial satellite company, while continuously tracking innovation directions in the aerospace field.

The turning point came in the second half of 2025, when Musk's Starlink achieved revenue exceeding 11.3 billion US dollars, with 60% of that as profit, validating a mature commercial closed loop. In December, Musk posted several consecutive posts on social media, with the core judgment that within the next 36 months, the cheapest place to deploy AI computing power will not be on Earth, but in space. The concept of space computing power immediately ignited markets at home and abroad.

This is an extremely grand vision that faces countless challenges both technically and in terms of cost. But Sun Xiaolei believes this direction is worth pursuing, stating that "the aerospace industry still needs people who dare to innovate."

More importantly, precisely because it is difficult, there are greater opportunities. "We are sufficiently professional in satellite development, and the AI track has a broad enough market space." After rigorous demonstration, Sun Xiaolei concluded that judging from the current technological accumulation and engineering feasibility, building a large-scale space computing power system has become achievable.

From its inception, Star Origin decided to focus its core business on space computing power satellites. Simply put, this means moving GPU computing power boards from ground data centers onto satellites, enabling the deployment and operation of large language models in space. Users on the ground can access on-orbit computing power and related services at any time via mobile phones or other terminals.

Star Origin's flagship product, "Xuanji No.1," is a 500-kilogram-class low-orbit satellite with a power output of 7 to 10 kilowatts, capable of carrying 9 to 16 computing power payloads. The preliminary design of the entire satellite has been completed.

According to Sun Xiaolei, over the past six months, leveraging the scientific research resources accumulated by the team at Harbin Institute of Technology and the Chinese Academy of Sciences, Star Origin has collaborated with leading industry enterprises to advance R&D, and has now made breakthroughs in key products such as high-computing-power payloads, low-cost large flexible solar arrays, and the active thermal control system combining fluid loops and deployable radiators.

"We want to build a data center in space," Sun Xiaolei said. "For future operations, we can draw on mature models from the ground. The current core priority is to first build the data center well and reduce costs."

The following is part of our interview with Sun Xiaolei, founder of Star Origin (edited):

36Kr: What does the BeiDou-3 experience represent?

Sun Xiaolei: It was China's first large-scale networking constellation. Before the BeiDou-3 system, domestic satellite engineering generally followed the traditional path of "single-satellite development and single-satellite deployment." BeiDou-3 achieved for the first time the integrated networking design and systematic construction of 30 satellites, breaking the previous single-satellite development paradigm.

During the construction of BeiDou-3, our team, with a relatively lean staffing, undertook the development of nearly half of the networking satellites.

With the implementation of a healthy competition mechanism in the industry, all organizations actively focused on technological iteration and process efficiency improvement, which not only reasonably optimized the overall cost of the entire satellite but also gradually cultivated a brand-new supporting supply chain covering all links with a very high degree of autonomy. This cohesion to take on tough challenges is the precious combat effectiveness the team formed back then.

36Kr: What exactly is the concept of the 10P computing power satellite you are developing? How is it different from other computing power satellites currently on the market?

Sun Xiaolei: After this satellite is launched into space, it does not need to transmit data back to the ground. It can run lightweight large models directly in orbit, which is not edge computing.

But this 10P is not a random number we pulled out of thin air. It is technically achievable at present, and we and our partners have completed the preliminary plan.

However, from 10P to 100P, there is a great deal of technology to overcome, such as flexible radiators, high-inertia satellite control, low-cost ultra-high-power flexible solar arrays, and satellite-ground laser communications, none of which are easy. We will first implement the 10P satellite, get the engineering process running, and then iterate upward.

36Kr: This sounds like a very long-term plan. What is the commercialization path at this stage?

Sun Xiaolei: We are currently drafting proposals for potential customers, striving to secure at least one order within this year.

What is the logic for customers to pay at this stage? After the computing power satellite is deployed in space, remote sensing data can be processed in real time in orbit without being transmitted back to the ground for calculation, which can form continuity with their previous investments.

Our planned pace is to strive for orders for high-computing-power satellites, with the goal of first getting the technology working, reducing costs in the future, and then conducting our own operations. The operational idea is to build a self-owned constellation, including hundreds of low-orbit computing power satellites that can later be expanded to thousands, to provide computing power services.

There is another issue that few people mention: even if we launch the computing power satellite, how do users access it? Therefore, we are also planning medium-orbit communication satellites. In the future, users can access the large models deployed in space by connecting through communication satellites.

36Kr: Constellation operation requires a large amount of capital to launch satellites independently. How big is the market space for space computing power? Can the costs be justified at present?

Sun Xiaolei: Current commercial rocket prices are generally 60,000 to 70,000 yuan per kilogram. After rocket reusability is realized, we hope to reduce this to 10,000 yuan per kilogram.

Satellite costs also need to be reduced. Currently, a satellite costs 30 million yuan, which needs to be reduced to around 10 million yuan through technological iteration.

We are building a calculation model — incorporating satellite costs and launch costs, then predicting the growth in AI computing power demand over the next five years to see if the ground can support it, and then inversely calculating the economic viability of space data centers. We will calculate domestic and international scenarios separately to assess the priority of market directions.

36Kr: China does not face the same energy crisis as the United States. Is it really more cost-effective to place data centers in space than on the ground?

Sun Xiaolei: No one could clearly explain how satellite internet would make a profit back then, but Musk made it happen. The same is true for space computing power. We believe the direction is correct, so we will move forward, and the path will become clear as we go.

Look at the explosive growth of AI in five years. Who could have predicted that five years ago? If it continues to develop for another five years with deeper penetration, computing power demand could be dozens or hundreds of times what it is now. At that point, the ground will not be able to support it, and space computing power will become a commercially viable solution.

There is also an unavoidable point: security. The Russia-Ukraine conflict has shown everyone how significant Starlink's role is. Ground data centers are the first to be bombed during wars. Moving them into space makes them out of reach for countries without sufficient capabilities. This is also why the state has been continuously promoting this direction.

36Kr: What is the operating model of space data centers? What will ultimately generate revenue?

Sun Xiaolei: We can refer to the model of ground data centers — whatever the ground does for operations, we will do the same in space. The billing method is also similar: either charge large model companies, or purchase our own models to provide to-C services. Fully independent operations may need to be planned around a 10-year development cycle.

The current key is not how to operate, but to build the "data center" itself well and drive down costs through technological iteration. Once costs drop to a certain level, operations will fall into place naturally.

Everyone knows that this requires early layout. Being two years late is not a big deal, but being 10 years late will be too late.