In 2025, Misha Laskin, a theoretical physicist and machine learning engineer, expressed concern over the state of artificial intelligence (AI) in the United States. Laskin, who contributed to the development of advanced AI models at Google, noted a significant rise in the use of open-source AI models, particularly those developed in China, which are gaining traction among American companies.

Laskin founded Reflection AI, a startup valued at $8 billion, to offer an American alternative to these increasingly competitive Chinese models. He observed that many of these models are not only cost-effective but also customizable, leading to their growing adoption in Silicon Valley.

Interviews with over 15 AI startup founders and industry experts revealed that while American AI models continue to lead in cutting-edge capabilities, many Chinese models are now sufficiently advanced for various applications and are often more affordable. This trend raises questions about the sustainability of the U.S. AI industry, which has heavily invested in companies like OpenAI and Anthropic, and whether the reliance on closed models may be misplaced.

Michael Fine, head of machine learning at Exa, noted that using Chinese models has proven to be faster and less expensive than relying on larger, proprietary models from American companies. Chinese models, such as DeepSeek's R1 and Alibaba's Qwen, are classified as open-source, allowing users to download and modify them freely, contrasting with the closed models from American firms that are accessed through proprietary systems.

Despite the historical dominance of American closed-source models, recent advancements by Chinese companies have narrowed the performance gap. Metrics from Artificial Analysis indicate that Chinese open-source models are now competitive with leading American models in several domains. This shift has led to platforms like OpenRouter witnessing a preference for Chinese models among users.

The Chinese government has actively promoted open-source technologies, with President Xi Jinping advocating for cooperation in this area. In contrast, the U.S. AI landscape has been largely shaped by private sector initiatives, with American companies maintaining a closed-model approach.

While some in Silicon Valley argue that American models still hold a significant edge in terms of capabilities and user convenience, concerns about the perceived risks associated with using Chinese models persist. A recent report from the U.S. Center for AI Standards and Innovation highlighted potential safety issues with popular Chinese models, while a White House memo raised concerns about Alibaba's alleged ties to the Chinese military.

As the competition intensifies, American AI companies and the federal government are recognizing the implications of the rise of Chinese models. Experts have warned that the lack of robust open-source models in the U.S. could pose an existential threat to democratic values. In response, initiatives such as the ATOM Project aim to revitalize American open-source efforts, emphasizing the need for the U.S. to maintain its leadership in the AI sector.

Recent developments include the White House's AI Action Plan, which encourages the promotion of open-source AI, and OpenAI's release of its first open-source model in five years. These efforts reflect a growing awareness of the importance of open-source models in fostering innovation and ensuring equitable access to AI technologies.