The Rise of Chinese Open-Source AI in the United States
Even as the United States and China engage in a fierce rivalry over the development and deployment of artificial intelligence, Chinese technology is quietly making significant inroads into the US market. Despite the ongoing geopolitical tensions, open-source AI models from China are gaining traction among a growing number of programmers and companies in the US.
These models differ from the closed generative AI systems that have become household names, such as OpenAI’s ChatGPT or Google’s Gemini. These proprietary models keep their inner workings closely guarded. In contrast, many Chinese open-source AI models, developed by companies like Alibaba and DeepSeek, allow developers to customize and adapt the software to meet specific needs.
According to a recent report by the developers’ platform OpenRouter and the US venture capital firm Andreessen Horowitz, the global usage of Chinese-developed open-source AI models has surged from just 1.2 percent in late 2024 to nearly 30 percent in August. This rapid growth highlights the increasing appeal of these models.
Wang Wen, dean of the Chongyang Institute for Financial Studies at Renmin University of China, noted that “Chinese open-source models are cheap — in some cases free — and they work well.” An American entrepreneur, who wished to remain anonymous, shared that their business saves $400,000 annually by using Alibaba’s Qwen AI models instead of proprietary alternatives.
“If you need cutting-edge capabilities, you go back to OpenAI, Anthropic or Google, but most applications don’t need that,” the entrepreneur said.
Several prominent US entities, including chip giant Nvidia, AI firm Perplexity, and Stanford University in California, are also incorporating Qwen models into their projects.
The DeepSeek Shock
The January launch of DeepSeek’s high-performance, low-cost, and open-source “R1” large language model (LLM) challenged the perception that the best AI technology must come from US-based giants like OpenAI, Anthropic, or Google. It also served as a wake-up call for the United States, highlighting how far China has advanced in the AI field.
In addition to DeepSeek, AI models from China’s MiniMax and Z.ai are gaining popularity globally. China is also entering the race to develop AI agents — programs that use chatbots to perform online tasks such as booking tickets or adding events to a calendar.
Agent-friendly and open-source models, like the latest version of Moonshot AI’s Kimi K2 model released in November, are seen as the next frontier in the generative AI revolution.
US Government Awareness
The US government is aware of the potential of open-source AI. In July, the Trump administration released an “AI Action Plan” that emphasized the need for “leading open models founded on American values.” The plan suggested that such models could become global standards.
However, so far, US companies have taken a different approach. Meta, which once led the US open-source movement with its Llama models, is now shifting focus towards closed-source AI. This summer, OpenAI — under pressure to revive its original nonprofit spirit — released two “open-weight” models, which are slightly less flexible than fully open-source models.
Building Trust
Among major Western companies, only France’s Mistral continues to support open-source AI. However, it lags significantly behind DeepSeek and Qwen in terms of usage.
Western open-source offerings are often viewed as less appealing. A US entrepreneur who uses Alibaba’s Qwen noted, “The Chinese open-source models are just not as interesting.”
Despite concerns about profitability, the Chinese government has actively encouraged the development of open-source AI technology. Mark Barton, chief technology officer at OMNIUX, mentioned that he is considering using Qwen but some of his clients may be hesitant to interact with Chinese-made AI, even for specific tasks.
Given the current US administration’s stance on Chinese tech companies, risks still exist. Barton told AFP, “We wouldn’t want to go all-in with one specific model provider, especially one that’s maybe not aligned with Western ideas.” He added, “If Alibaba were to get sanctioned or usage was effectively blacklisted, we don’t want to get caught in that trap.”
However, Paul Triolo, a partner at DGA-Albright Stonebridge Group, stated there are no “salient issues” surrounding data security. “Companies can choose to use the models and build on them…without any connection to China,” he explained.
A recent Stanford study highlighted that “the very nature of open-model releases enables better scrutiny” of the technology. Gao Fei, chief technology officer at Chinese AI wellness platform BOK Health, agrees. “The transparency and sharing nature of open source are themselves the best ways to build trust,” he said.
