US-China AI Geopolitical Competition: Proprietary Frontier LLMs vs. Open-Source Models

Abstract

This research paper presents an expanded, comprehensive analysis of the global geopolitical competition in artificial intelligence (AI) between the United States of America and the People's Republic of China (PRC) amidst the structural realignment of the international order. The central thesis investigates the strategic, architectural, and ideological confrontation between U.S. proprietary (closed) frontier Large Language Models (LLMs) and China’s counter-strategy of scaling and diffusing high-performance Open-Source AI architectures. Grounded in structural neorealism, techno-nationalism, and weaponized interdependence theories, this study examines how algorithm ownership, compute infrastructure, and software diffusion reshape the global distribution of power.

The paper demonstrates that in response to Washington’s "Small yard, high fence" doctrine, Bureau of Industry and Security (BIS) export controls, and semiconductor chokepoints (NVIDIA, ASML, TSMC), Beijing has deployed an asymmetric counter-strategy. By disseminating state-of-the-art open-weights models (e.g., DeepSeek-R1/V3, Alibaba Qwen-2.5) across the Global South, China disrupts Western digital monopolies, institutionalizes its technical standards, and fosters long-term hardware-software ecosystem dependencies. Utilizing comparative case studies, geoeconomic modeling, and qualitative institutional analysis, the article evaluates the acceleration of digital fragmentation ("Splinternet") and the dilemmas of digital sovereignty facing developing nations. Finally, the study formulates a strategic "Digital Multi-vectorism" policy framework for middle powers, with a dedicated case study on the Republic of Kazakhstan’s national AI development and diplomatic positioning in a technological bipolar world.