China's Open-Weights AI Strategy Is Outpacing the US
Original: China's open-weights AI strategy is winning
Why This Matters
China's open-weights approach could reshape global AI ecosystem dominance as US frontier model advantages narrow.
China's strategy of releasing open-weights AI models is gaining global traction, according to Ben Werdmuller writing on werd.io. With US export controls limiting China's compute access, Chinese firms turned this into a distribution advantage. An a16z partner noted an 80% chance any given startup now uses Chinese AI models.
Ben Werdmuller argues that China's decision to release AI models as open-weights has proven strategically effective, turning a US-imposed compute disadvantage into a distribution advantage. Because AI models themselves have minimal technical moat — users can switch between ChatGPT and Claude with minimal workflow disruption, and API-based engineers can swap providers even more easily — the real value lies in enterprise services and ecosystem lock-in, not the models alone.
US export controls on GPUs have limited China's ability to build OpenAI-scale centralized services, but open-weights models sidestep this barrier entirely: they are portable, permissionless, and can be hosted and customized freely. This mirrors the historical pattern where open infrastructure tends to dominate adoption.
Meanwhile, the performance gap is closing. Moonshot and Alibaba have released models claimed to match OpenAI and Anthropic at a fraction of the cost. A16z partner Martin Casado noted in The Economist that there is now an 80% chance any given startup is using Chinese AI models. Werdmuller warns that the US closed-first approach — historically at odds with its own open-internet strategy — risks ceding AI ecosystem leadership to China across manufacturing, scientific research, and enterprise sectors globally.