China's Open-Weights AI Strategy Is Winning

What happened

A viral analysis (900+ HackerNews points) argues that China is systematically winning the AI race by releasing high-quality open-weight models for global adoption — while America's closed, proprietary approach is losing the economic foundation of AI deployment.

Context and impact

AI models lack lasting moats beyond brand loyalty. Open models enable permissionless global adoption. US export controls paradoxically disadvantage American companies by preventing centralized service deployment, while Chinese models rapidly close performance gaps. The economic logic underpinning massive US AI infrastructure investments ($300B+ Stargate) could collapse as inference costs fall.

Details

• Economic thesis: inference costs fall → closed models lose price advantage
• China released Kimi K3 (2.8T params), Qwen 3.8 (2.4T), DeepSeek V4 — open-weight or freely accessible
• US export controls prevent American firms from global deployment, not Chinese ones
• HN community (900+ points): majority agrees with the open-model thesis
• Author warns: $300B Stargate investment logic assumes a lasting price moat