China's open-weights AI strategy is proving successful, with its companies gaining a competitive edge. In contrast, America's closed, proprietary approach to AI development is seen as a losing strategy that could impact the US economy. AI models themselves have limited long-term technical moats, with the real value lying in surrounding enterprise services. Users can easily switch between models and vendors based on performance, especially when accessed via API.

The US government's export controls on GPUs and data regulations limit Chinese companies' ability to offer global-scale centralized services. However, open technologies typically lead in infrastructure adoption due to their permissionless nature, allowing for greater innovation, customization, and local hosting. China's open-weights models leverage this, turning a compute disadvantage into a distribution advantage and commoditizing the layer where US companies profit.

While US frontier models have historically outperformed open ones, this gap is narrowing. Companies like Moonshot and Alibaba are releasing models comparable to those from OpenAI and Anthropic at lower costs. This shift, coupled with the observation that 80% of startups may be using Chinese models, suggests a significant change in AI leadership. The article argues that America's closed-door policy for a technology with minimal moats is a flawed strategy, advocating for a more open and collaborative approach to foster ecosystem benefits and align incentives for long-term success.