China’s state-developed DeepSeek AI is significantly accelerating the design of advanced combat aircraft, changing the dynamics of military AI competition.
The integration of large-scale language models (LLMs) into aircraft construction marks a decisive turning point for the Chinese aviation industry. The Shenyang Aircraft Design Institute uses DeepSeek models in the development of the J-15 and J-35 fighter jets, making design processes significantly more efficient. The AI takes over time-consuming tasks such as aerodynamic simulations and stress tests, which significantly shortens innovation cycles.
Particularly noteworthy is the hybrid mixture-of-experts (MoE) architecture of the DeepSeek models, which enables cost-efficient performance. The upcoming DeepSeek-R2 model will be equipped with 1.2 trillion parameters and will have 97.3% lower training costs than comparable Western models.
Military applications beyond construction
DeepSeek technology is already being used in PLA hospitals to analyze combat injuries and optimize medical logistics. Experts expect it to soon expand to real-time surveillance analysis, where AI could correlate satellite imagery, drone footage and electronic signals to identify enemy assets.
This technological advance reflects China’s strategy to circumvent US semiconductor restrictions through software innovation. By optimizing algorithms for efficiency, the Shenyang Institute can run advanced simulations on domestic GPUs such as the Biren BR100, avoiding reliance on banned NVIDIA chips.
Ads
Executive Summary
- DeepSeek AI is actively used in the development of China’s J-15 and J-35 fighter jets
- The technology halves development cycles for flight control algorithms and other critical components
- DeepSeek-R1 can analyze 10,000 sensor data points from wind tunnel tests in under 3 seconds
- AI optimizes stealth technologies and suggests new material compositions
- China’s use of AI in military technology reduces dependence on Western technologies
- This development intensifies the global AI race in the defense sector
Source: SCMP