appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Dannie Peng

The global race to deploy data centres in space to power artificial intelligence (AI) development has begun, with China seizing an early lead.
“In terms of the highest performance of space-based computing or the completeness of [the] technological framework, we are at the forefront,” Beijing-based computer scientist Han Yinhe said.
Han is a researcher at the Institute of Computing Technology (ICT) of the Chinese Academy of Sciences in Beijing, and is taking part in a major initiative to develop a real “space supercomputer” in low-Earth orbit that will integrate 10,000 high-performance computing cards.
But the United States is now rushing to catch up with much more money and launch capabilities.
Tech billionaires in Silicon Valley, including Elon Musk and Google CEO Sundar Pichai, have been drawing up plans to move AI infrastructure into space.
On Monday, Musk wrote on social media that he was envisioning plans to launch a million tonnes of satellites into space annually, with the aim of deploying 100 gigawatts of AI computing power in orbit each year. Last month, he confirmed that SpaceX planned to build orbital data centres by scaling up its next-generation Starlink V3 satellites.
Just days earlier, Pichai announced that Google planned to launch space-based data centres as early as 2027.
“We’ll send tiny, tiny racks of machines and have them in satellites, test them out and then start scaling from there,” he said. The technology giant unveiled its initiative called Project Suncatcher in November.
The push to deploy AI into orbit is a strategic gambit with far-reaching consequences for national security, economic dominance and the future of global connectivity such as 6G. Orbital data centres offer unlimited solar power, natural heat dissipation and near-instant data processing, cutting reliance on ground infrastructure and bandwidth.
The country that wins the race to build scalable, reliable and powerful space-based AI infrastructure will gain a decisive edge in the next era of computing – much like the US did with the internet and China did with 5G.


One American company taking action is Starcloud. In early November, the start-up launched Starcloud-1, a refrigerator-sized satellite carrying the first Nvidia H100 graphics processing unit into orbit.
Boasting 80 gigabytes of graphics memory and performance far surpassing that of any previous spaceborne computer, this chip will conduct in-orbit testing of Earth observation image analysis and large language model workloads.
While many take it as the start of a space AI race, it is not the first to do so.
Beijing-based aerospace company Zhongke Tiansuan, an ICT spin-off, launched a space computer equipped with home-grown high-performance chips as early as 2022, which has operated stably in orbit for more than 1,000 days to date.
Last year, the team launched an upgraded system, achieving in-orbit AI inference ahead of Starcloud. These endeavours have enabled China to accumulate longer in-orbit data and expertise in engineering.
“China has already taken the lead in engineering implementation speed and commercialisation,” said Zhao Hongjie, vice-president of Guoxing Aerospace, an AI satellite developer based in Chengdu, in an interview with the state-owned Science and Technology Daily last month.
In May, Guoxing Aerospace launched a cluster of 12 satellites in collaboration with Zhejiang Lab, a research institute backed by the Zhejiang provincial government.
This formed the foundation of the world’s first orbital computing constellation, one that offers a combined computing power of 5 peta operations per second and carries a space-based AI model with 8 billion parameters.
In September, this in-orbit constellation showed its commercial applications by conducting calculations for the transport sector in space.

Processing remote sensing imagery of Pazhou district in Guangzhou, the cluster completed both inference and result transmission in just three minutes, significantly reducing response times and saving over 90 per cent of ground transmission bandwidth.
Zhao said that while Western nations were accelerating their deployment efforts, no large-scale computing constellations comparable to this one had yet been launched into orbit and put into commercial operation. The firm unveiled its second constellation project in October.
At the end of November, Beijing municipal authorities also unveiled a plan to build a massive, centralised space data centre in a dawn-dusk orbit 700-800km (435-497 miles) above Earth. Led by an innovation consortium, the project targets a system with a power capacity exceeding 1GW.
According to its developer, deployment is expected to be completed in three phases, culminating in a megawatt-scale orbital data centre by 2035.
As AI fuels an insatiable appetite for computing power, the idea of orbital data centres is being embraced enthusiastically, as the facilities can use unlimited solar power and radiate their heat into space, reducing energy needs.
Traditionally, satellites collect data in space, but then have to send it back to Earth for processing. Yet on-orbit data centres enable low latency, near real-time analytics and autonomous operations. This is particularly important for situations such as disaster monitoring and military reconnaissance.
But taking high-performance computing systems into space involved much more than simply sending chips up, noted Han from the ICT.




Existing computing chips face severe challenges when deployed in space, such as exposure to high-energy particles, extreme temperature fluctuations and intense vibrations during rocket launches, all of which threaten their stable operation.
Han said that since 2019, some domestic institutes including ICT had initiated research projects exploring how to deploy AI chips in space while ensuring their reliability.
State news agency Xinhua reported on December 5 that the Tiansuan team had managed to address radiation-induced failures such as computational errors and system crashes in industrial-grade chips. They were also experimenting with fluid-loop cooling in orbit to transfer heat from high heat-flux chips.
Han also said that without access to the most powerful chips from the US, China was seeking to develop more innovative architecture and hardware design.
“We are devoted to integrating multiple home-grown high-performance chips onto a single satellite and designing an optimal architecture to enhance capability,” he said, adding that this was about systemic innovation rather than mere stacking.
He also revealed that his team was developing an advanced next-generation computing system that was set for in-orbit trials next year.
However, industry experts noted that space data centres still faced a series of core challenges. Beyond stringent technical requirements, there are also cost bottlenecks in building large-scale constellations in the future.
Han said that, much like what happened in terrestrial transport, the construction of more data centres in orbit would encourage Chinese commercial space companies to refine their technologies and reduce costs over time.
“Ultimately, this will become a profitable business,” he added.


China tested a handheld submarine detector. It tracked a subway train underground

How Wang Zhenyi, Chinese female astronomer and mathematician, defied gender expectations

Chinese drug loberamisal extends stroke treatment window from 6 hours to 2 days

China-Russia space camp in Malaysia opens to youth in Southeast Asia

Science

The race for AI centres in space: why the US goes all in after China makes first move

The push to deploy AI into orbit is a strategic gambit with far-reaching consequences for national security and economic dominance. Photo: Shutterstock

China and the US are in a race to be the first to launch AI space centres into orbit, with far-reaching consequences. Photo: Shutterstock

US seeks ‘constructive’ reset with China as trade tensions ease: Greer

US Trade Representative Jamieson Greer during an interview at the White House in October. Photo: Reuters

AI race comes down to power and data centres – and China has the edge, says unicorn hunter

Chinese venture capitalist Allen Zhu Xiaohu of GSR Ventures Management. Photo: Handout

A report published in June by US think tank Rand said the surge in computing demand was creating “exceptional” pressure on America’s power grid. Photo: Shutterstock

Over 10 years in the making: China launches 2,000km-wide AI computing hub

Chinese scientists say their innovation means computing centres separated by distance but linked by optical network could work together almost as efficiently as a single giant computer. Image: Shutterstock

China’s geopolitical competition expands into new frontiers with scientific milestones

Members of a Chinese scientific expedition team deploy ice-based buoys in the high-latitude waters of the Arctic Ocean on August 25. Photo: Xinhua

China’s Tan Suo San Hao (Explorer Three) polar research vessel has successfully conducted a scientific expedition performing crewed dives in the Arctic Ocean. Photo: CCTV

News

China

Beijing and Washington battle for tech domination with next-gen plans to launch supercomputers into orbit and transform global connectivity


Beijing and Washington battle for tech domination with next-gen plans to launch supercomputers into orbit and transform global connectivity.


The race for AI centres in space: why the US goes all in after China makes first move

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Beijing and Washington battle for tech domination with next-gen plans to launch supercomputers into orbit and transform global connectivity

Beijing and Washington battle for tech domination with next-gen plans to launch supercomputers into orbit and transform global connectivity