[UK Think Tank: The Next 'China Shock' Will Hit Defense]

According to an article published by the Royal United Services Institute (RUSI) on August 21, 2026: If future wars are primarily fought by machines, victory will belong to the nation capable of integrating artificial intelligence with robotics and mass-producing them at scale—today, that nation is China.

If most combat in the next conventional war is conducted by unmanned systems, military power will be redefined as the ability to fuse AI with robotic technology and manufacture it on a massive scale.

Anyone seeking to understand the form of the next conventional war should observe what’s happening at the headquarters of San Francisco’s AI unicorns and in Shenzhen’s factories—not at Washington’s and Moscow’s power palaces. For Europe, the starting point is rather awkward. In the foreseeable future, Europe cannot produce world-class frontier AI models. Meanwhile, in high-end industrial production, Germany is being surpassed by China—a trend equally evident in electric vehicles and defense sectors.

——The Rise of Machines

China is training robots for combat in dedicated centers, and its lack of real-world combat experience may be purchased. In 2024, China installed 295,000 industrial robots, compared to 32,200 in the U.S. and 26,982 in Germany. Once the global leader in industrial robotics, German-based KUKA, was acquired by China in 2016. In humanoid robotics, Chinese companies captured nearly 90% of the global market share by 2025. Leaders Unitree Robotics and AgiBot each sold over 5,000 robots, while total sales across all U.S. manufacturers fell short of 500 units.

Western robots have material cost lists nearly three times higher. True, most military robots won’t engage in direct combat but will perform reconnaissance, mine clearance, and supply transport. Even combat-capable robots won’t be humanoids—they’ll be autonomous quadrupeds, drones, and vessels. But the key point is that they all rely on similar autonomous system manufacturing capabilities, and their bills of materials may even include identical actuators, batteries, and sensors.

——China’s Industrial Dominance

China’s leadership is no accident. It mirrors the strategy that propelled China to dominate the electric vehicle market: early state support, dozens of competitors entering the field, and direct flow of components into the defense supply chain. An analysis by War on the Rocks describes an implicit administrative ecosystem composed of incentive mechanisms, certification systems, and procurement platforms—systematically transforming commercial tech firms into suppliers for the People’s Liberation Army. At this year’s World Defense Show in Riyadh, a company linked to the Chinese government unveiled a mass-produced robotic dog equipped with four anti-tank missiles, suitable for export.

Behind all this lies the hardest bottleneck to overcome. China produces 90% to 98% of the rare-earth permanent magnets used in every drone motor and every actuator—and since 2025 has been tightening export controls on rare earths. German auto parts suppliers understand the implications: magnet shipments stalled almost brought production lines to a halt, halting even Tesla’s Optimus robot manufacturing. Defense equipment is no exception. Over 80,000 components across approximately 1,900 U.S. weapon systems directly depend on minerals controlled by Beijing. Every drone Rheinmetall, Helsing, or Stark wants to mass-produce requires a magnet whose supply Beijing could cut off at any moment.

What about artificial intelligence? According to Stanford University’s AI Index, the performance gap between Chinese and American models has largely closed—while private AI investment in China remains under 5% of that in the U.S. Part of the reason for this rapid catch-up lies in straightforward replication: Chinese models are trained using outputs from U.S. cutting-edge models, thereby inheriting their capabilities at extremely low cost. America’s AI edge is being “distilled” away, as Chinese models are effectively “attending classes” taught by American ones.

Moreover, large-scale model training and operation ultimately depend on electricity and chips. China excels here too: in 2024, China added about 429 million kilowatts of new power generation capacity, compared to just 51 million kW in the U.S. and around 20 million kW in Germany. Since Chinese models can’t access the latest NVIDIA chips, they’ve been optimized for extreme efficiency—a feature crucial in resource-constrained battlefield computing environments.

Differing from leading models at U.S. firms like OpenAI and Anthropic, China’s top-tier large models are open-weighted and priced up to 50 times cheaper than their American counterparts. This explains their immense popularity: Alibaba’s Qwen series has already surpassed one billion downloads.

When we transplant this landscape into the military domain, enormous proliferation risks emerge: open-weight models are not subject to export controls—although the Chinese government is now considering imposing such restrictions. They aren’t confined to running only in U.S.-controlled hyperscale data centers, lack mandatory security filters or built-in government “kill switches,” and their open weights can be adjusted and optimized by both state and non-state actors for specific purposes. Furthermore, Chinese models have never gained access to the newest generation of NVIDIA chips. It is therefore reasonable to infer they will perform exceptionally well in resource-limited battlefield environments—what technologists call “hardened edge computing.”

——Looking Beyond China

What does this mean for Europe and the Western alliance? Minerals, metals, and magnets are now strategic assets. So too are training data generated by autonomous systems in the field. China spent decades building an industrial base that systematically pushed European and American industries out of one global market after another. To assume these capabilities would remain idle during worsening geopolitical tensions would be naive. According to a widely cited estimate, China’s civilian drone factories could produce nearly one billion armed drones annually if given orders.

Disclaimer: The above report excerpt is from a RUSI column article.

[Follow this Toutiao account for more military news]

Original source: toutiao.com/article/1875005023759370/

Disclaimer: The views expressed in this article are those of the author alone.