StarX Future has successively completed two rounds of financing worth hundreds of millions of yuan, with 37 GPUs adopting advanced manufacturing processes operating stably in orbit, and the company is deploying the first cloud-collaborative space computing constellation, the "Fuyao Project".
Text by | A Zhi
On September 11, at the special forum for innovative development of space computing power under the China Computing Power Conference, Xingce Future officially announced the completion of two consecutive rounds of financing, A+ and A++, with a total amount of hundreds of millions of RMB, jointly invested by well-known industrial and financial investment institutions including Bei Venture Capital, China Creation Ventures, SunShine Capital, NewDing Capital, while existing shareholders Wuxi Venture Capital and Houtian Capital continued to increase their investment.
Hundreds of millions of RMB raised within half a year is more than three times the total financing amount of 5 rounds that Xingce Future secured in the past 5 years.
This scene is more like an emotional epitome of the entire track.
Elon Musk broke the valuation ceiling of SpaceX with the logic of "AI + Space", and even stated that AI may account for 99% of the company's value in the next four or five years. In his narrative, SpaceX is a technology company that takes Starship as infrastructure and AI as its core value, and in the future, the main battlefield for AI inference and computing will be in space. Space computing power has instantly become the hottest track of global hard technology in 2026, with extremely high value in the primary market.
In China, this narrative has been quickly absorbed and amplified. A large number of entrepreneurs have poured into the track, and capital flows continuously.
Standing in this wave, Xingce Future has been rediscovered by the market.
This team, originating from the "TianGe Project" of Tsinghua University, proposed on-orbit AI and satellite intelligence at the very beginning of its entrepreneurship. But that was 6 years ago, when the term "space computing power" had not even started to circulate.
In retrospect, Xingce Future is the earliest enterprise in China that has proactively laid out space computing power and achieved a closed commercial loop, leading the industry in both standardized mass delivery scale and commercial implementation speed: 37 GPUs with advanced manufacturing processes have been deployed in orbit, 39 payloads are operating stably in orbit, the company has taken the lead in realizing the on-orbit application of 12nm advanced process GPUs in China and 7nm advanced process GPUs globally, with a single device operating stably in orbit for more than four years.
But this is not a story of a lucky man catching the windfall. As Cang Jirong, CEO of Xingce Future, said, "We have almost been fighting against the wind all the time, and it is not until this year that we start to catch the favorable wind."
The background of this statement is not complicated. In 2020, when the company was founded, the first round of financing boom in commercial aerospace had passed, the industry was still in the stage of route exploration and infrastructure acceleration, the demand of the application market had not yet risen, and the epidemic further aggravated the external complexity that start-ups had to face.
Perhaps because it was born in the "cold winter", the team got used to lurking, focused on R&D and on-orbit verification, and counted every penny carefully. While many space computing power players were still demonstrating product reliability and queuing for space launch, Xingce Future had already completed the first stage of computing payload verification and commercial application.
"Space computing power should not be a concept that only stays on PPT, but a landed on-orbit productivity. There are various stories and technical routes in the market, all of which are nothing but empty talks if they have never been launched into space. In the end, we have to see whether there is real on-orbit operation and whether there are market payers." Judging from the number of stable on-orbit operations and engineering experience, Xingce Future stepped out of the industry concept stage earlier, and also fulfilled the future it bet on 6 years ago.
But the core of this story is not the financing amount and data, but how a long-underrated team seizes its own opportunities and places bets for the next decade.
Entry: Targeting AI + Aerospace, the atypical entrepreneurship of three "Tsinghua alumni with all degrees from Tsinghua"
The origin of Xingce Future's entrepreneurship can be traced back to 2016. At that time, Cang Jirong and Cao Dezhi, who were still pursuing their doctoral degrees at Tsinghua University, were both core R&D backbones of the "TianGe Project".
This is a deep space scientific research project led by Tsinghua students. They independently developed scientific payloads, carried them on satellite platforms into space, with the main goal of detecting and locating gamma-ray bursts in the universe. In layman's terms, it is to monitor the short moment of "cosmic fireworks" in space, then quickly "tell" the ground the message, and then the ground telescopes will conduct joint observations for further scientific research.
There are two key variables here: one is the ability of payloads to identify and judge independently, and the other is to transmit conclusions as quickly as possible.
In October 2018, the first payload of the "TianGe Project" successfully entered orbit, completing the full-process on-orbit verification of the world's first gamma-ray burst detection payload for micro-nano satellites — this is also the source of Xingce Future's on-board intelligent processing technology and engineering capability accumulation.
For Cang Jirong, this experience made him form a key judgment: only when satellites have the ability to independently process, filter and make decisions on data in orbit, can they maximize the value of their applications. Satellites need to be intelligent, and AI is the core driving force.
In 2020, Cang Jirong and Cao Dezhi, who had just graduated with their doctoral degrees, hit it off and founded Xingce Future; at the end of the same year, Pan Xingyu, another Tsinghua PhD, joined. The three "Tsinghua alumni with all degrees from Tsinghua" (referring to people who completed their bachelor's, master's and doctoral studies at Tsinghua University) officially formed the co-founding team, and started technology and industrial exploration in the cross field of "AI + Aerospace".
With on-board AI computing as the core, Xingce Future officially launched its first space-based computing intelligent payload "Xingxi" in 2021.
The team reached a consensus and chose the right direction, but during the epidemic, it was not easy to solve the capital problem. Almost all the first angel round of funds was invested in R&D, but the first computing payload had to wait for launch verification to get real market feedback. At the most tense stage of cash flow, there was only money left for two or three months of operation on the company's account. In order to support the operation of the company, the team began to take some external technical service orders.
"These orders are not large in scale, but they kept the company alive and also completed our first lesson in commercialization." Cang Jirong does not shy away from talking about this difficult experience, and what he repeatedly emphasizes is the other side: when the team moves up from the trough, it should lurk when it needs to, accumulate technology and engineering capabilities first, respect the market, and understand operation.
In February 2022, the launch verification window finally arrived. The first computing payload of Xingce Future was equipped with a 12nm chip, successfully realized on-orbit operation, and has been working ever since. The company's second round of financing also arrived smoothly, and the feedback from customers and market further boosted the team's confidence.
One detail is that after 2022, no one in Xingce Future's core team resigned no matter how big the challenge was.
This group of doctoral students from Tsinghua University has no background in aerospace research institutes. Standing at the intersection of aerospace and AI, relying on scientific research capabilities, engineering experience and rapid iteration, they took the lead in sending commercial advanced process chips into space in China, and took a route that was not "favored" in the past.
"Aerospace is a systematic project. Why can the cost of Starlink be reduced? It is because a large number of commercial off-the-shelf (COTS) chips are used. There must be a lot of trial and error in this process, find problems and solve them, and what matters is the iteration speed." In the 3 hours of our communication, Cang Jirong mentioned "iteration" 17 times — which is more in line with the thinking mode of the Internet and embodied intelligence. He believes that iteration efficiency can be transformed into product competitiveness, and organizational form directly affects decision-making efficiency.
Based on this cognition, in addition to forming a core R&D team, Xingce Future has been continuously absorbing professional talents from the fields of AI, aerospace, Internet, autonomous driving and even large-scale infrastructure, and has built a business-oriented, efficiency-first operation and management system, becoming one of the few atypical entrepreneurial teams with mixed backgrounds in the commercial aerospace field.
Lurking: "Leveling up by fighting monsters" in space, engineering experience is the real barrier
If you want to summarize Xingce Future's direction choice on the technical route in one sentence, the conclusion can be drawn first: the same as Elon Musk.
The core contradiction of space computing power at present is: aerospace-grade chips are naturally reliable, but they have high cost, slow iteration and closed ecology; ground advanced process chips have strong performance, mature ecology, and their performance is increased by an order of magnitude per generation, but they cannot work directly in the space environment.
Xingce Future chose the latter: based on the mainstream GPU ecology on the ground, it solves the reliability problem through system fault tolerance and space protection design, and takes the route of low cost, high performance, fast iteration and wide compatibility.
Three layers of guarantee are made in the specific scheme design:
The first layer is structural shielding, by thickening the structural parts, the cumulative radiation damage is reduced to less than one thousandth of the original;
The second layer is system fault tolerance, using ECC check (a memory error correction technology), redundant backup, watchdog and other methods to deal with transient errors such as single event upset;
The third layer is active avoidance, integrating a small radiation sensor on the payload to detect the particle threat level in real time, and triggering different levels of self-check, repair or even power-off protection according to the threat level.
The third layer of design is Xingce Future's exclusive capability, based on the accumulated space environment cognition of the Department of Engineering Physics of Tsinghua University in the past.
"In the past, everyone had similar concepts of structural shielding and triple modular redundancy, but it is another thing to achieve it in engineering. We have one more layer of space environment perception, one more dimension of physical information." Cang Jirong said.
To understand the difference between the two routes in a popular way, the past idea of aerospace to solve radiation resistance is to "resist rigidly", using special process chips to fight radiation at the physical level. Xingce Future's idea is to allow chips to make mistakes, make the system smart enough to independently correct errors, quickly identify and repair, without affecting the completion of on-orbit tasks.
This design scheme is not conceived in the laboratory, but an engineering capability continuously precipitated through long-term stable on-orbit operation.
After the first computing payload of Xingce Future was successfully verified in orbit, market demand followed, product verification accelerated, and the real "leveling up by fighting monsters" began.
In the orbit of 500 to 630 kilometers, the team once encountered the situation that high-energy particles overturned the storage device and the hard disk could not be recognized, and finally recovered by re-uploading the software; they also encountered current abnormalities and functional failures caused by single events, and realized minute-level task recovery through power-off restart, watchdog reset and other methods.
Cang Jirong mentioned that the team's real cognition of space radiation was obtained when errors occurred, "When no errors occur, we even doubt whether it is necessary to make so many redundant designs. Until we experience problems, we find that many previous designs are valuable."
From taking the lead in realizing the on-orbit application of domestic 12nm chips in 2022 to the world's first 7nm chip launched into space in 2024, a counter-intuitive fact is that in the matter of on-board intelligent processing, domestic enterprises are not backward, and even go further globally.
But in Cang Jirong's view, computing power entering orbit does not mean that it can operate in orbit, and on-orbit operation does not mean that it can really generate value. In actual market applications, business indicators such as boot time, task volume, number of processed pixels per unit time, and Token throughput (TPS) are more important than nominal computing power — which also corresponds to the first item in Xingce Future's corporate values: value creation.
"Xingce Future is a commercial company. Scientific research innovation and technological breakthroughs are means rather than ends. We need to continuously create value for employees, shareholders, the industry and the country." Cang Jirong said.
Reflected in commercial behaviors, Xingce Future is not a company that sells hardware payloads, and it does not do "one-off deals". After the satellite is launched, the delivery team will process data, adjust algorithms, and help users run services at the front line of users. Xingce Future began to participate in the national-level remote sensing constellation project in 2023, successfully completed the delivery, and then successively established stable cooperation with the "national team" units in the aerospace field and leading commercial constellation operators.
In the past few years, Xingce Future has actually stood behind satellite companies, and jointly explored the value of satellite applications with industry users through hardware and software integrated products + services.
Nowadays, this set of strategies is extending to a longer industrial chain: computing payload is the starting point, scenario service is the entry point, and computing power operation is the end point.
Correspondingly, Xingce Future's business logic is also carried out in three layers: the first layer is supporting computing payload hardware, which now has the capability of mass delivery and on-orbit deployment; the second layer is the sensing and computing integrated satellite application service for vertical industries, which has landed cases in emergency, transportation and other fields; the third layer is space computing power infrastructure and Tokenized operation services, which is a longer-term layout direction.
Three Growth Curves of Xingce Future's Business
The market feedback from the application level is the most direct.
In the field of emergency energy, Xingce Future serves State Grid Corporation of China, providing an intelligent computing and algorithm acceleration platform for the power infrared satellite A, realizing real-time on-orbit processing of infrared remote sensing data, automatic fire point identification and high-precision positioning. The payload was put into orbit in February 2026. Official data shows that the detection and positioning time of a single scene image is less than 100 milliseconds.
In terms of waterway transportation, Xingce Future cooperates with CCCC Water Transportation Consultants Co., Ltd., using satellite remote sensing and on-orbit computing power to monitor the congestion situation of waterways, and expand scenarios such as major engineering monitoring, multi-dimensional environment monitoring, and global AIS tracking enhancement.
After the implementation of key industry applications, Xingce Future targeted three core scenarios of financial insurance, energy monitoring, and emergency disaster relief, and reached in-depth strategic cooperation with relevant leading units; at the international level, it reached strategic cooperation with Hainan Satellite Data and Application Research Center, to jointly build space computing public technology products represented by "international intelligent remote sensing satellite constellation, space-ground integrated cross-border sensing and computing platform", helping to build an international open cooperation ecosystem for space-based intelligent applications.
A set of latest data shows that Xingce Future's current on-hand order scale has exceeded 100 million RMB, the company's commercialization process is accelerating, and its revenue scale is expected to double in 2026.
In terms of delivery guarantee, the company has reserved more than 100 sets of computing platforms to be launched, which can support up to 10POPS computing power, and has formed a computing power product matrix covering "on-board edge intelligence" to "space computing power center".
Leap: With the "Fuyao Project" as the core, what kind of company does Xingce Future want to be?
In the past 6 years, Xingce Future's development layout can be summarized in four words: compute space data in space.
Letting the data generated in space be calculated in space depends on three premises: the hardware is reliable enough, the on-orbit data is sufficient enough, and the software ecology is gradually improved. Only when all three are available can large and small models be deployed in orbit and scheduled collaboratively, and the space computing power center can be possible to move from concept to engineering.
Following the path of "actual measurement — verification — large-scale implementation", Xingce Future has run through the technical system and commercial closed loop of computing space data in space, and is simultaneously laying out the large-scale networking and commercial services of space computing power clusters. On this basis, it promotes the evolution from computing space data in space to computing ground data