China has built and launched a super-powered artificial intelligence (AI) system designed to tap directly into its national supercomputing network and conduct high-level scientific research on its own.
The scheme was launched on December 23, a month after US President Donald Trump unveiled the Genesis Mission – an “AI Manhattan Project” aimed at securing US technological dominance.
While the US plan is bound by tight timelines, requiring proof of progress within 270 days, China has already deployed its system at scale, serving more than a thousand potential institutional users nationwide.
The new AI agent can take simple natural language commands and then independently break down complex research tasks, allocate computing power, run simulations, analyse data and generate full scientific reports – with no need for constant human oversight, according to the official China Science Daily.
The system runs on China’s National Supercomputing Network (SCNet), a high-speed digital backbone linking over 30 supercomputing centres across the country.
It supports nearly 100 scientific workflows in fields such as materials science, biotechnology and industrial AI.
What once took scientists a full day can now be done in about an hour, according to the SCNet.
This network could reshape the course of the US-China tech rivalry and have wider implications globally.
For decades, both nations have invested billions of US dollars in supercomputers capable of simulating nuclear explosions, designing next-generation stealth fighters and discovering life-saving drugs.
A critical question in the tech world has been whether AI could be granted access to such sensitive, often classified data, transforming it into a “super scientist”.
Giving AI direct access to and control over national supercomputing networks could dramatically speed up scientific breakthroughs – but this also carries unprecedented risks.
These include leaking data held by the government about its citizens, opening the way for hackers to collect sensitive information or giving AI access to details about weapons systems.
“Science is shifting from number crunching to AI-powered discovery. Such new AI agents will connect the tools, data and computing power scattered across different systems, giving scientists better tools to innovate faster,” Qian Depei, a member of the Chinese Academy of Sciences and head of an SCNet expert panel, told the press conference launching the project.
“AI for Science is not only a technical pathway but also a transformation in how research is organised,” Cao Zhennan, deputy director of the High Performance Computer Research Centre, told the event.
In August, Beijing launched its national “AI+” initiative, explicitly calling for AI to be used to accelerate scientific discovery.
The Genesis Mission was Washington’s answer, with plans to use federal supercomputers and decades of government research data to train powerful AI agents.
Launching the project, Trump said it was designed to “invest in AI-enabled science to accelerate scientific advancement” and further “America’s technological dominance and global strategic leadership”.
Led by the Department of Energy, the mission will establish an American Science and Security Platform, which will use the US government’s scientific data sets – the world’s largest collection – to train AI models and create intelligent agents.
These agents will be used to test new hypotheses, automate research workflows and speed up scientific discovery.
The platform will use massive government supercomputers to train AI models and create intelligent agents. These agents will test hypotheses, automate research tasks and speed up breakthroughs.
The US plan has a strict timeline. Within 60 days of launching, it must identify 20 national level science and technology challenges the mission could address, and must demonstrate operational capability within 270 days.
Six priority areas have been specified – advanced manufacturing, biotechnology, critical materials, nuclear fission and fusion energy, quantum information science, and semiconductors and microelectronics.
China’s SCNet was launched in 2023 and aims to integrate supercomputing and intelligent computing resources across the country, linking supercomputing centres in a high-speed network that supports national development.
The platform was officially launched in the Binhai Hi-Tech Zone in Tianjin in April 2024 and has since connected over 30 computing centres, which serve users from more than a thousand government agencies, enterprises, universities and research institutions. – South China Morning Post
