Former Google China head Kai-Fu Lee is making major adjustments to his artificial intelligence startup 01.AI strategy. The core of this adjustment lies in the full adoption of Deepseek's open source model. Kai-Fu Lee believes that this move poses a survival challenge to OpenAI's business model. In an interview with the South China Morning Post, Kai-Fu Lee elaborated on the background and significance of this strategic change in detail.
Kai-Fu Lee revealed that 01.AI has abandoned its previous strategy of training proprietary large language models and relied entirely on Deepseek's open source products. He said the release of Deepseek triggered a "ChatGPT moment" in China, driving the integration of domestic hardware and software providers with the Deepseek model. The decision came after Chinese companies surged in demand for the Deepseek model at the end of January this year. Kai-Fu Lee believes that Deepseek's free and open source method poses a fundamental challenge to OpenAI.
Kai-Fu Lee further pointed out that Deepseek's free model poses a huge threat to OpenAI. "The biggest nightmare for Sam Altman is that his competitors are free. I've met a lot of people who canceled ChatGPT subscription because Deepseek is free," he said. This view reflects the strong advantage of the open source model in market competition.
With 200 employees, 01.AI now plans to focus on customizing Deepseek models for corporate clients in the fields of finance, gaming and law. Despite the strategic shift, Lee insists that his company's technical expertise remains valuable, especially in model training, adjustment and reasoning. He expects revenue to reach RMB 100 million (US$13 million) in the first quarter of 2025, equivalent to the company's full-year revenue in 2024, but 01. AI has not yet achieved profitability.
Regarding OpenAI and Anthropic's recent move to ban the Deepseek model, Kai-Fu Lee interpreted it as a "pretense of paranoidness." "They watched the house of cards they built start to collapse because someone built an equally good house for free," he said. This comment reveals the impact of the open source model on traditional business models.
Kai-Fu Lee also pointed out the huge gap in operating costs between the two companies: OpenAI reportedly operates at $7 billion in 2024, while Deepseek only needs about 2% of that amount. He described Deepseek as "unlimited forever" because its founder has enough funds to maintain operations and reduces calculation costs by 5 to 10 times. This comparison further highlights the advantages of open source models in cost control.