Hugging Face CEO: Why don't American companies develop open-source AI?

Title

China is Leading the Way in Open Source AI, and Hugging Face CEO Wants to Know Why American Companies Aren’t Following.

Summary

Hugging Face co-founder and CEO Clément Delangue said he didn’t expect American startups and tech giants to be so inactive—China’s lead in open source AI is widening, while there is little movement in the U.S. Models like Qwen and DeepSeek from China have higher download and actual usage rates on Hugging Face compared to American models. Companies like Airbnb and Pinterest have already started using these Chinese models for a simple reason: they are cheaper and easier to modify. This gap is noteworthy: the U.S. has ample funding and leading computing power, but these advantages are not reflected in the adoption of open source models.

Analysis

Delangue oversees one of the largest open source AI platforms, and he has a clear view of the data. By early 2026, the proportion of Chinese models used by startups on Hugging Face is expected to exceed 30%. The reason is not complicated: Chinese labs default to open source, while American companies typically stick to closed source.

Looking deeper, there are more interesting points. American VC investment in AI is about 12 times that of China, and there are export bans on chips. But when it comes to actual usage data, Chinese models are still winning. About 80% of startups choose models like Qwen because they can cut costs by 5 to 10 times and can create custom models that closed-source ones cannot.

For American companies willing to open up, the opportunity is there. The AI framework from the White House in 2026 mainly discusses standards and safety, with little mention of openness. Another statistic: the proportion of Chinese employees regularly using AI tools is 83%, while in the U.S. it is 65%—the gap is not just about whether models can be accessed.

Impact Assessment

• Importance: High
• Category: Open Source, Industry Trends, Market Impact