From "Connected" to "Understood", Space-Based Communication and Computing Integration Startup Bets on Unmanned Systems, Secures Tens of Millions of Yuan in Series A Financing | 36Kr Exclusive

Text | A Zhe

Cover Source | Pexels

The large-scale deployment of China's low-orbit satellite constellation has entered a critical time window.

A set of intuitive progress data shows that: China SatNet has completed its first-generation network deployment, the number of in-orbit satellites of Yuanxin Satellite has exceeded 200, the Long March 10B rocket has successfully completed the world's first maritime mesh recovery, and the leading commercial companies are also approaching the key maiden flight milestone for their reusable rockets.

As the cost of "reaching space" is expected to drop rapidly, the focus of the entire industrial chain is shifting from "how to launch satellites" to a more pragmatic question: After the satellite network is built, who will it serve and how will it be used?

There is no standard answer to this question yet, but the market is full of players eager to provide solutions.

A startup company established less than half a year ago, focusing on "communication and computing integration for unmanned systems" 「Starlink Tianshu」, recently officially announced the completion of tens of millions of yuan in angel round financing, which is also the first round of financing completed by Starlink Tianshu.

In simple terms, the core business of Starlink Tianshu is to enable unmanned systems such as drones, autonomous vehicles, and unmanned ships to connect directly to satellites through an extremely lightweight satellite communication terminal, complete AI processing and decision-making on the collected data on the satellite, and then transmit the results back to the ground — This "ground data processing and space computing" model is not just about selling equipment, but about bundling communication and computing power to provide integrated communication and computing service solutions for unmanned systems.

This is also a signal. It means that during the transition period when China's low-orbit constellation moves from experimental verification to large-scale commercial use, entrepreneurs in the industrial chain have begun to dive into more specific application scenarios, to explore the key proposition of how to monetize the capabilities of space infrastructure.

Why now? Why unmanned systems?

The core team of Starlink Tianshu comes from the 54th Research Institute of China Electronics Technology Group Corporation, the 8th Academy of Aerospace and the Chinese Academy of Sciences. The resume of Cheng Yinghui, the founder, has a distinct "national team" background. From 2013 to 2024, he served at the 54th Research Institute of China Electronics Technology Group Corporation, successively holding positions as chief designer, deputy chief engineer, and chief engineer of the satellite communication system, deeply participating in key projects such as the low-orbit satellite internet project and a certain high-orbit satellite communication system.

The core factor driving Cheng Yinghui to start his own business is the arrival of the industrial inflection point — China's rocket technology and satellite platform integration capabilities are continuously improving, satellite internet has moved from a niche track in the past to the global spotlight, and the evolution path of Starlink has also led him to believe that "In the next stage, China will usher in a major explosion in the direction of satellite communication and integrated communication and computing."

And the question of "Why choose to start from unmanned systems" points to a deeper industrial judgment.

Cheng Yinghui defines unmanned systems broadly, covering various platforms such as drones, autonomous driving vehicles, unmanned ships, and embodied robots. He believes that, In the next 3-5 years, these unmanned systems from different fields will be the core incremental user group of satellite internet.

The demand is urgent.

On the one hand, unmanned systems have rigid requirements for networks, and ground networks are difficult to meet their needs in many scenarios. For example, when a drone flies hundreds of meters high, the ground network connection will be basically interrupted, and ocean-going unmanned ships are almost communication isolated islands. In these scenarios, satellite internet is an excellent solution.

Therefore, since transmitting data to space is the optimal solution, the choice between "only providing communication" and "integrated communication and computing" has become very clear. Satellite data traffic is inherently more expensive than ground traffic. Compared with transmitting all raw data back to the ground for processing, completing data processing on the satellite and returning decision results results in lower traffic costs and lower latency.

"Unmanned systems will be the first type of users that urgently need to 'go to space' to process data, which is also our original intention of proposing integrated communication and computing." Cheng Yinghui said. Of course, the future of unmanned system communication does not require users to choose between satellites and other methods, but will move towards a space-ground integrated three-dimensional network, which intelligently selects the optimal link based on factors such as mission requirements, environment, and cost budget.

However, to be frank, this direction is still in the early stage of development.

The two major domestic low-orbit constellations have not yet been officially put into commercial use, the pricing system for data traffic has not been established, and the on-board intelligent computing capabilities still need to be verified. This means that neither the large-scale shipment of terminal equipment nor the successful operation of the business model of integrated communication and computing services can be achieved without the further maturity of upstream infrastructure.

Cheng Yinghui has a relatively calm judgment on the timeline, "The real explosion of the market will take three years at the fastest, or five years at the slowest. But by that time, we will have truly developed the ability to achieve self-sustaining growth."

Divergence of Paths and Practical Choices for Space Computing Power

If we break down the long-term goal of realizing "integrated communication and computing", what Starlink Tianshu focuses on at this stage is actually two things: First, to enable unmanned systems to stably connect directly to satellites; second, to efficiently process specific data from unmanned systems on satellites.

These two goals correspond to its two core products at the current stage: the intelligent satellite terminal for unmanned systems, and the on-board intelligent computing payload for unmanned systems.

On the end side, the product should be made as lightweight as possible. Especially in aviation scenarios, the optimization of weight and volume directly corresponds to the real usage and decision-making threshold of customers. The first intelligent satellite terminal for unmanned systems developed by Starlink Tianshu weighs 2kg, which is reportedly the most lightweight Ka-band broadband satellite communication terminal in China at present.

The technical difficulty and R&D cycle of the satellite-side product are obviously more "challenging". What is worth paying attention to is the differentiation of its architecture route selection.

Different from the scheme of "stacking GPU computing power", Starlink Tianshu adopts a heterogeneous computing architecture. The logic behind this is not to pursue the absolute peak of computing power, but to derive the technical route and hardware design from the data characteristics of unmanned systems based on the required scenarios.

"Unmanned systems carry too many types of payloads — optics, infrared, SAR, LiDAR, and various data types are different. If all are processed by GPU, the efficiency is very low, and the cost and power consumption are high." Cheng Yinghui mentioned, "We are targeting the earliest data that needs to be processed in space to define the product, so the hardware architecture needs to adapt to different types of data." Under the realistic conditions of limited on-board power consumption and heat dissipation capabilities, this adaptive design is more feasible in engineering.

Only when the deployment on the end side and the satellite side is further scaled up, can data truly collaborate to form an "integrated communication and computing" solution. From product verification to large-scale expansion, this process will take about 3 to 5 years.

For Starlink Tianshu, the two most important things this year are to accelerate the launch of self-developed satellite terminal products, with the first batch of prototypes expected to be launched in September. At present, it has reached cooperation with a domestic emergency management drone project, a new energy heavy-duty truck enterprise, and a transportation group; it will complete the construction of the ground "unmanned system ground data and space computing" verification system within the year, first run the entire process on the ground, and then advance to the space-ground joint experiment stage next year, to send the computing power payload into space for in-orbit verification.

A long-term goal is that Cheng Yinghui hopes Starlink Tianshu can become a leading space-based communication intelligent service provider in the field of unmanned systems.

The choice of this direction, to some extent, reflects the divergence of paths in the domestic space computing power track.

Currently, domestic companies that bet on space computing power can be roughly divided into two categories. One type of players want to directly build space-based data centers, and their path is closer to "first move computing power to space, then gradually reduce costs", which is aligned with the long-term logic of Starship + Starlink. The prerequisite for its realization is to break through key bottlenecks such as rocket launch costs, satellite energy systems, heat dissipation technology, and chip radiation resistance one by one. The other type is the path chosen by Starlink Tianshu, which does not directly compete with ground computing power on cost, but first forms a closed loop in rigid scenarios such as unmanned systems, and then expands to a larger ecosystem.

Behind the two paths are the differences in the judgment of the industrialization rhythm of space computing power.

In Cheng Yinghui's view, It will take a long time for the cost of space computing power to catch up with ground computing power. The advantage of focusing on communication and computing services for unmanned systems is that the commercial closed loop can be achieved faster. At least after the satellite network is initially deployed, we can first solve the problem of who to serve and how to serve.

Looking further into the future, they are also trying to answer a bigger question: When satellite internet is no longer a scarce resource, what will the application layer on top of space infrastructure look like.

For a startup company established less than half a year ago, it is too early to talk about the final outcome. But at least, it has provided a solution that is different from "moving data centers to space". And from having the right direction to achieving commercial success, Starlink Tianshu still needs to use product implementation, customer verification, and in-orbit data to gradually bridge the path that all hard technology companies cannot avoid.