Chinese entrepreneur Kai-Fu Lee anticipates that only three AI models will endure the domestic upheaval in China, catalyzed by the emergence of DeepSeek, which has revolutionized the existing paradigm for artificial intelligence development. Lee, a venture investor and founder of 01.AI, is placing his bets on DeepSeek, Alibaba Group Holding Ltd., and ByteDance Ltd. to dominate the Chinese market. In the US, he foresees four leading models: xAI, OpenAI, Google, and Anthropic.

DeepSeek, a model purportedly trained at a significantly lower cost than OpenAI’s, has expedited global development since its introduction in January. The Chinese startup showcased the efficacy of the open-source framework, prompting more developers to embrace this approach. This is partly due to US sanctions aimed at restricting the most advanced Nvidia Corp. chips from entering China, compelling local developers to innovate.

Lee argues that these sanctions have not been effective, stating that DeepSeek has actually thrived with fewer resources, enabling training and inference at a fraction of the cost incurred by OpenAI and others. Lee’s company, which initially focused on building large AI platforms, has shifted its strategy away from pursuing massive, trillion-parameter models due to their high capital intensity. Instead, it has pivoted towards partnerships with Alibaba and DeepSeek to develop industry-specific applications, aiming to construct what Lee terms the operating layer atop DeepSeek’s underlying technology.

Lee questions the sustainability of OpenAI’s model, given its $7 billion operating costs in 2024, compared to DeepSeek’s significantly lower expenses. He suggests that with such a formidable competitor, OpenAI’s Sam Altman may be losing sleep. Venture capitalists in both the US and China are increasingly shunning funding for pioneer models, opting instead to invest in AI infrastructure, application companies, and enterprise and consumer apps. In China, local governments are intervening to bolster AI development after Beijing designated the sector a priority.

Despite the shift in funding focus, Lee notes that AI funding remains robust, with investors simply redirecting their attention away from pioneer models.