Investing.com -- Kai-Fu Lee, CEO of 01.AI, has questioned the sustainability of OpenAI’s business model in a recent interview with Bloomberg’s The China Show. Lee expressed his support for DeepSeek, a new AI model from China that he believes could pose a significant challenge to OpenAI’s ChatGPT.
Lee likened DeepSeek’s release in China to ChatGPT’s U.S. release, suggesting that DeepSeek has already become a cultural cornerstone in the country. Many CEOs are eager to implement DeepSeek into their companies, viewing it as an exceptional model that can connect to corporate databases and deliver value at little to no cost.
Lee explained his support for DeepSeek by pointing to the consolidation of pretraining in open-source models, underscoring his belief that open-source models will shape the future of AI over closed-source alternatives.
The CEO then questioned why OpenAI and Anthropic have not yet conceded in the AI race, given that they have built their businesses on the premise of creating closed models intended to outperform any open-source models. According to Lee, these companies were caught off guard when they encountered an open-source model like DeepSeek, which matches their capabilities but comes at a much lower cost.
Lee emphasized OpenAI’s high operating costs, suggesting that DeepSeek developed its models using just 2% of the resources OpenAI expends. He argued that the key issue isn’t whose model is marginally better, but whether closed-source businesses have a sustainable plan.
OpenAI reportedly spends $7-8 billion annually, resulting in a "massive loss," especially as DeepSeek offers a comparably capable model for free. Lee described DeepSeek as "infinitely lasting," noting that its founder, Liang Wenfeng, has sufficient funding to sustain it at its current level and has reduced computation costs by a factor of 5 to 10.
According to Lee, the emergence of such a formidable competitor capable of slashing costs significantly is likely causing OpenAI CEO Sam Altman to "lose sleep."
Regarding the future of AI in China, Lee believes that only 3 models will prove sustainable: DeepSeek, Alibaba (NYSE:BABA)’s Qwen, and potential models from ByteDance. Notably, he did not mention Chinese startup Moonshot’s Manus AI, a recent release that has been gaining global attention.
Lee clarified that it’s not AI itself that’s becoming commoditized, but rather the foundational models on which AI continues to build. He also highlighted a trend among venture capitalists, who are shifting their focus from funding pioneering models to investing in AI infrastructure and applications.
Despite his prediction that only 7-8 winners will emerge in the AI industry between China and the U.S., Lee maintains that the field remains highly competitive, with many players continuing to vie for dominance.