The article argues that China's strategy of releasing open-weights AI models is outperforming the locked-down, proprietary approach of American AI companies. This trend suggests that open-source AI development is gaining competitive advantage.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
While U.S. AI companies cling to closed, proprietary models, China is gaining an edge in the global AI race through open-weight strategies. This strategic shift is not only reshaping the tech ecosystem but could also have profound economic consequences for the United States.
U.S. AI companies (e.g., OpenAI and Anthropic) adopt closed, proprietary models, while Chinese companies tend to release open-weight models. This strategic difference stems from different market incentives: U.S. companies are forced to pursue direct profits, whereas Chinese companies gain ecosystem benefits through openness.
The U.S. government has imposed export controls on GPUs and restricted data sharing with Chinese servers. As a result, Chinese companies cannot offer global-scale centralized services, but by releasing open-weight models, they turn their compute disadvantage into a distribution advantage—models can be freely hosted, modified, and adapted, gaining wider adoption worldwide.
AI models as products have almost no moat; users can easily switch between ChatGPT and Claude with minimal impact on workflows. In engineering, accessing models via API requires only changing the API endpoint to use the same prompts.
Although companies can lock in customers through contracts, the technical switching costs are low in the long run. Users will choose the model that best fits their needs and switch providers when another model becomes better. This dynamic makes it difficult for closed strategies to retain customer loyalty.
U.S. frontier models have historically led open models, but the gap is narrowing. Moonshot and Alibaba have released models claiming to rival the best from OpenAI and Anthropic, at lower cost. These rapid releases suggest that the U.S. lead in AI frontiers is becoming increasingly tenuous.
a16z partner Martin Casado noted in The Economist that 80% of startups are using Chinese models, and Chinese models are poised to take the lead. This further confirms the effectiveness of China's open strategy.
Open technologies often win in infrastructure adoption because they can be used without permission, becoming centers for further innovation. Open-weight models, while not fully open-source, are portable and permissionless, allowing users to freely host, experiment, modify, and adapt them.
By releasing open-weight models, China enables easy integration across industries such as manufacturing and scientific research, gaining significant ecosystem benefits. In contrast, the closed strategies of U.S. companies limit the formation of such ecosystems.
The U.S. economy is currently largely driven by AI spending. If the AI spending bubble bursts—which is likely given the dynamics above—the consequences could be severe. U.S. companies need to adjust their incentive structures, shifting from pursuing direct profits to focusing more on ecosystem benefits.
The author calls for a more nuanced U.S. strategy that supports public AI, federal services, and open research, allowing open technologies to operate in the public interest and under public values.
This article is based on an analysis piece by author Ben Werdmuller, citing reports from The Verge and comments from a16z partner Martin Casado in The Economist. Claims about Chinese model performance (e.g., 'rivaling OpenAI and Anthropic') are source claims and have not been independently verified. The percentage of startups using Chinese models (80%) is also a source claim.
Through its open-weight AI strategy, China is exploiting the weaknesses of the U.S. closed model, gaining a global ecosystem advantage. If the U.S. persists with its closed strategy, it risks an AI spending bubble burst and economic consequences.
Primary report
Primary source