At a time when authorities in both nations have urged their military to be war-ready, the debut of the Norinco P60 was hailed by Communist Party officials in news releases as an early demonstration of how Beijing is employing DeepSeek and AI to catch up in its arms race with the United States. A Reuters analysis of hundreds of research papers, patents, and purchase records provides a glimpse of Beijing’s methodical attempts to use AI for military purposes.
Although the precise workings of China’s next-generation weapon systems and the extent of their deployment are state secrets, procurement records and patents provide insights into Beijing’s advancement toward capabilities similar to those of the United States, such as autonomous target recognition and real-time battlefield decision support. Patents may not always represent functional technology, and Reuters was unable to determine whether all of the devices had been manufactured. According to the papers, tenders, and patents, the PLA and its affiliates are still using and searching for Nvidia chips, including those that are subject to U.S. export restrictions.
Since the documents don’t specify when the gear was exported, Reuters was unable to ascertain whether the chips were stocked prior to Washington’s prohibitions. Military-affiliated research institutes have been using them, according to patents issued as recently as June. In September 2022, the U.S. Commerce Department restricted exports to China of Nvidia’s popular A100 and H100 processors. While the company is unable to monitor individual resales of previously sold products, Nvidia spokesman John Rizzo told Reuters that “recycling small quantities of old, second-hand products doesn’t enable anything new or raise any national security concern.” It would be impossible to use limited equipment for military purposes without maintenance, software, or support. Questions concerning Reuters’ findings were not answered by the US Treasury or Commerce Department.
In 2025, the Chinese military has also expanded its use of contractors who say they only use domestically made hardware, such as Huawei AI chips, according to Sunny Cheung, a fellow at the Jamestown Foundation defense policy think tank in Washington, who examined hundreds of PLA Procurement Network tenders over the course of six months this year. Although Reuters was unable to independently verify his claim, the change would take place at the same time as Beijing is publicly pressuring domestic companies to use technology produced in China.
The demand for and usage of Huawei processors by PLA affiliates was discovered by the news agency through its study of procurement notices and patents submitted to China’s patent office. However, it was unable to confirm all of the tenders that Jamestown, which is publishing a story this week that it sent to Reuters ahead of time, saw. When questioned about the use of its semiconductors in military applications, Huawei remained silent. Requests for comment regarding DeepSeek, Norinco, and the Chinese defense ministry’s use of AI for military purposes were not answered. Similar inquiries were not answered by the academic institutions and defense companies that submitted the research papers and patents that Reuters was able to view. DEEPSEEK INDEPENDENCE
Twelve PLA entities’ tenders that Reuters was able to view this year included the use of DeepSeek models, but just one of them mentioned Alibaba’s Qwen, a significant domestic competitor. A request for comment regarding the military’s usage of Qwen was not answered by Alibaba. Jamestown claims that during 2025, procurement announcements pertaining to DeepSeek have increased and that fresh military applications are frequently posted on the PLA network. The PLA’s preference for DeepSeek also underscores China’s goal of “algorithmic sovereignty,” which Beijing refers to as lowering reliance on Western technology while enhancing control over vital digital infrastructure. Regarding the PLA’s employment of AI, the US Department of Defense chose not to comment.
“DeepSeek has voluntarily provided, and is likely to continue to provide, support to China’s military and intelligence operations,” a State Department spokeswoman told Reuters in response to questions. The spokesperson went on to say that Washington will “adopt a bold, inclusive strategy to American AI technology with trusted foreign countries around the world, while keeping the technology out of our adversaries’ hands.” AI-BASED APPLICATIONS AND PLANNING According to the documents, China is considering visually immersive command centers, sophisticated war game simulations, AI-powered robot dogs that scout in packs, and drone swarms that follow targets on their own. A sci-fi-style call for AI-powered robot dogs that would patrol jointly for threats and clear hazardous dangers was issued by the PLA in November 2024.
Reuters was unable to determine whether the tender was completed. Images published in state media show that China has previously used armed robot canines from AI robotics developer Unitree in military exercises. Questions concerning Unitree’s PLA work were not answered. The examination of academic papers, patents, and tenders issued in the last two years demonstrates how PLA and related organizations are using AI to enhance military planning, including creating tools to swiftly analyze satellite and drone photos.
In a white paper published in February to advertise their services, researchers at Landship Information Technology, a Chinese company that incorporates AI systems into military vehicles like Norinco’s, claimed that their technology, which is based on Huawei chips, can quickly identify targets from satellite imagery while coordinating with radars and aircraft to carry out operations. According to Xi’an Technological University, AI has also reduced the time it takes military planners to move from locating and designating a target to carrying out an operation. In a May overview of their findings, researchers at that institute claimed that their DeepSeek-powered system could evaluate 10,000 battlefield scenarios in 48 seconds, each with its own unique set of variables, topography, and troop deployments.
They said that a traditional team of military strategists would have needed 48 hours to finish such a mission. The researchers’ claims could not be independently confirmed by Reuters. AUTOMATIC WEAPONS According to the documents, Chinese military organizations are spending money on combat equipment that is becoming more autonomous. The military is trying to incorporate artificial intelligence (AI) into drones so they can identify and track targets and cooperate in formations with minimal human involvement, according to two dozen of the tenders and patents obtained by Reuters. According to a patent application this year, Beihang University, which is well-known for its research on military aviation, is utilizing DeepSeek to enhance drone swarm decision-making when aiming for “low, slow, small” threats—the military abbreviation for drones and light aircraft.
Amid mounting concerns that an unfettered deployment of AI-powered bombs could result from a dispute between Beijing and Washington, Chinese defense chiefs have made a public commitment to retaining human control over weapons systems. In an effort to offset China’s numerical advantage in unmanned aerial vehicles, the U.S. military, which is also investing in AI, plans to deploy thousands of autonomous drones by the end of 2025, according to authorities. Chinese and American models In marketing materials, Chinese defense contractors such as Shanxi 100 Trust Information Technology have emphasized their dependence on locally made parts, such as Huawei’s Ascend CPUs, which enable AI models to function. Questions concerning the company’s ties to the PLA and Huawei were not answered.
According to a survey of patent filings spanning the last two years, military-affiliated academics continue to often use Nvidia technology in their studies despite the shift to domestic CPUs. 35 applications were found by Reuters that mentioned the usage of Nvidia’s A100 chips by scholars at the National University of Defense Technology (NUDT) of the PLA and the “Seven Sons,” a consortium of Chinese universities that are subject to U.S. sanctions and have a track record of carrying out defense-related research for Beijing. During the same time period, those organizations submitted 15 patent applications mentioning Huawei Ascend hardware, which was created as an alternative to Nvidia chips.
The PLA Rocket Force University of Engineering recently submitted a separate patent application in June for a remote-sensing target identification system that allegedly trained models using A100 chips. U.S. restrictions have affected their AI development “to some degree,” according to Senior Col. Zhu Qichao, who heads a NUDT research center, who told Reuters last year that they are committed to closing the technology gap. By claiming that China “has more than enough domestic chips for all of its military applications,” Nvidia’s Rizzo downplayed PLA desire for the company’s technology.
