China’s Military Deploys Cost-Efficient DeepSeek AI Across Drone Swarms And Robot Dogs

Photo credit: Unitree

Amazon Deals: Forget the DJI Mini 4K. Get a Skyrover X1 Fly More Combo for $718!

China’s People’s Liberation Army has rapidly integrated DeepSeek artificial intelligence into military weapons systems ranging from autonomous combat vehicles to drone swarms, according to a Reuters investigation that reviewed hundreds of research papers, patents, and procurement records. The development threatens to reshape the balance of autonomous warfare capabilities between Beijing and Washington.

The investigation reveals systematic Chinese military adoption of DeepSeek—the AI startup that shocked Silicon Valley in January 2025 by matching U.S. competitors’ performance at a fraction of the cost. While American policymakers spent years restricting chip exports to slow China’s AI progress, Beijing responded by developing more efficient systems that work around hardware limitations.

Norinco Unveils DeepSeek-Powered Combat Vehicle

In February 2025, China’s state-owned defense giant Norinco showcased the P60 military vehicle capable of autonomously conducting combat-support operations at speeds up to 50 kilometers per hour (31 mph). Communist Party officials touted the P60’s release as an early demonstration of how Beijing is using DeepSeek AI to close the gap with the United States in the accelerating arms race between the two superpowers.

The P60 represents more than just another autonomous vehicle. Its DeepSeek-powered brain signals a fundamental shift in how China approaches military AI development—prioritizing efficiency and cost-effectiveness over brute-force computing power. This strategy mirrors DeepSeek’s civilian breakthrough that trained its R1 model for approximately $6 million, compared to the $100 million-plus that U.S. companies typically spend on comparable AI systems.

Researchers at Landship Information Technology, which integrates AI systems into military vehicles including Norinco’s platforms, claimed in a February white paper that their DeepSeek-based technology built on Huawei chips “can rapidly identify targets from satellite imagery, while coordinating with radars and aircraft to execute operations.”

PLA Procurement Records Show DeepSeek Dominance

Usage of DeepSeek models appeared in a dozen procurement tenders from PLA entities filed in 2025, while only one referenced Alibaba’s Qwen, a major domestic AI rival, according to analysis by the Jamestown Foundation. DeepSeek-related procurement notices have accelerated throughout 2025, with new military applications appearing regularly on the PLA Procurement Network.

The AI model’s popularity with China’s military reflects what Beijing calls “algorithmic sovereignty”—reducing dependence on Western technology while strengthening control over critical digital infrastructure. This push intensified after the U.S. Commerce Department banned exports to China of Nvidia’s popular A100 and H100 chips in September 2022, citing national security concerns.

Alibaba declined to comment about military use of its Qwen AI system. The Chinese defense ministry, DeepSeek, and Norinco did not respond to Reuters’ requests for comment about their use of AI for military applications.

Robot Dogs and Autonomous Drone Swarms

The PLA issued a tender in November 2024 for AI-powered robot dogs that would scout in packs for threats and clear explosive hazards. While Reuters could not confirm whether the tender was fulfilled, China has previously deployed armed robot dogs from manufacturer Unitree in military drills, according to state media images.

Unitree’s robot dogs were spotted climbing stairs during a March 2025 demonstration in Hangzhou, Zhejiang province, showcasing the rapid pace of development. The four-legged platforms combine mobility advantages in complex terrain with AI-powered visual search technology for reconnaissance and strike missions.

Beihang University, renowned for military aviation research, is using DeepSeek to improve drone swarm decision-making when targeting “low, slow, small” threats—military shorthand for drones and light aircraft—according to a 2025 patent filing. Two dozen tenders and patents reviewed by Reuters show Chinese military efforts to integrate AI into drones for autonomous target recognition, tracking, and formation flying with minimal human intervention.

AI Cuts Battlefield Planning From 48 Hours to 48 Seconds

The practical implications of DeepSeek integration extend beyond individual weapons platforms. Xi’an Technological University researchers claimed in a May 2025 summary that their DeepSeek-powered system assessed 10,000 battlefield scenarios—each with different variables, terrain, and force deployments—in just 48 seconds. “Such a task would have taken a conventional team of military planners 48 hours to complete,” they stated.

Reuters noted it could not independently verify the researchers’ claims. However, the assertions align with DeepSeek’s demonstrated capabilities in complex problem-solving and strategic planning scenarios that require analyzing multiple variables simultaneously.

The time compression represents a significant tactical advantage. Military operations increasingly depend on the speed of the “kill chain”—the process from target identification to strike execution. AI systems that can analyze satellite imagery, coordinate with radar networks, and recommend courses of action in seconds rather than hours fundamentally change battlefield tempo.

Working Around U.S. Chip Restrictions

Despite U.S. export controls aimed at limiting China’s access to advanced AI chips, the PLA continues using Nvidia hardware while simultaneously pivoting toward domestic alternatives. Reuters identified 35 patent applications from the PLA’s National University of Defense Technology and affiliated “Seven Sons” universities referencing Nvidia’s A100 chips. Those entities filed 15 additional patents citing Huawei Ascend hardware, designed as a domestic substitute for Nvidia processors.

Patents filed as recently as June 2025 show PLA Rocket Force University of Engineering using A100 chips for model training in a remote-sensing target detection system. How Chinese military researchers obtained these chips after the September 2022 export ban remains unclear. Reuters could not determine if the chips were stockpiled before Washington imposed restrictions.

Nvidia spokesman John Rizzo told Reuters that while the company cannot track individual resales of previously sold products, “recycling small quantities of old, second-hand products doesn’t enable anything new or raise any national security concern. Using restricted products for military applications would be a non-starter, without support, software, or maintenance.” Rizzo added that China “has more than enough domestic chips for all of its military applications.”

Chinese Military Pivots to Huawei Chips

The Chinese military has increased its use of contractors claiming to exclusively use domestically made hardware like Huawei AI chips in 2025, according to Sunny Cheung, a fellow at the Washington-based Jamestown Foundation who analyzed several hundred tenders from the PLA Procurement Network over six months this year.

Chinese defense contractors like Shanxi 100 Trust Information Technology have promoted their reliance on domestically produced components like Huawei’s Ascend chips in marketing materials. The shift would coincide with a public pressure campaign by Beijing encouraging domestic firms to adopt China-made technology.

Huawei declined to comment about military deployment of its chips. The company faces its own restrictions from Washington, having been placed on the U.S. Entity List in 2019 over national security concerns.

Human Control Commitments Amid Autonomy Push

Chinese defense leaders have publicly committed to maintaining human control over weapons systems amid growing concern that a conflict between Beijing and Washington could lead to unchecked deployment of AI-powered munitions. However, the line between human oversight and autonomous operation continues to blur as AI systems become more sophisticated.

Chinese military entities are investing in increasingly autonomous battlefield technology, procurement documents suggest. Two dozen tenders and patents show efforts to integrate AI into drones for target recognition and tracking, as well as formation flying with little human intervention.

The U.S. military faces similar questions about autonomous weapons as it races to deploy thousands of AI-enabled drones by the end of 2025. Pentagon officials frame the push as necessary to counter China’s numerical advantage in unmanned aerial vehicles.

U.S. State Department Responds

A State Department spokesperson told Reuters that “DeepSeek has willingly provided, and will likely continue to provide, support to China’s military and intelligence operations.” The spokesperson added that Washington will “pursue a bold, inclusive strategy to American AI technology with trusted foreign countries around the world, while keeping the technology out of the hands of our adversaries.”

The U.S. Department of Defense declined to comment about the PLA’s use of AI. The Treasury Department and Commerce Department did not respond to questions about Reuters’ findings.

Senior Colonel Zhu Qichao, who leads a research center at the National University of Defense Technology, told Reuters last year that U.S. restrictions have impacted Chinese AI research “to some degree,” though researchers remain determined to narrow the technological gap.

DroneXL’s Take

The DeepSeek story represents a fascinating twist in the tech Cold War between Washington and Beijing. For years, U.S. policymakers believed chip export controls would slow China’s AI development. Instead, those restrictions forced Chinese researchers to innovate around constraints—and they succeeded spectacularly.

We’ve been tracking this trend closely at DroneXL. Just two days ago, we reported on China’s National Youth Intelligent Unmanned Systems competition, where over 200 university teams—including students who built an autonomous patrol-and-strike robot dog—competed in military drone scenarios organized by China’s defense leadership. That wasn’t an isolated academic exercise. It was talent recruitment for exactly the kind of AI-powered autonomous weapons systems now deploying across the PLA.

The pattern extends beyond ground robots. We covered China’s massive stealth fighter drone unveiling in September, noting that Beijing appears to be leapfrogging the U.S. “loyal wingman” concept and moving directly to fully autonomous air superiority fighters. In June, we examined how Chinese military factories are using robotic arms to manufacture drones with precision that exceeds human technicians—drilling holes twice as fast with zero quality rejects.

Meanwhile, the Pentagon’s efforts stumble forward. Our September investigation into the Replicator program revealed widespread technical failures, with drone boats going adrift due to steering problems and target-recognition software failing during military drills. The program has been transferred to a new Defense Autonomous Warfare Group in a bid to accelerate progress before a critical 2027 deadline.

Here’s what drone professionals need to understand: DeepSeek’s military applications won’t stay confined to battlefields. The same cost-effective AI techniques that let China train advanced models on weaker chips will eventually filter into commercial drone systems. If Chinese manufacturers can build autonomous targeting systems that assess 10,000 scenarios in 48 seconds, imagine what that computational efficiency means for obstacle avoidance, flight planning, and autonomous operations in civilian applications.

The irony is profound. Washington’s chip restrictions were supposed to maintain America’s AI advantage. Instead, they catalyzed development of more efficient AI that doesn’t need cutting-edge hardware—exactly the kind of innovation that makes advanced capabilities accessible to nations without access to Nvidia’s latest chips.

For the drone industry, this development signals a fundamental shift. The next generation of autonomous systems won’t necessarily come from whoever has the most powerful processors. They’ll come from whoever uses available computing power most cleverly. China just demonstrated they’ve mastered that approach—and the PLA is already deploying it at scale.

What do you think? Share your thoughts in the comments below.