Kai-Fu Lee, former head of Google China, is pivoting his AI startup, 01.AI, to fully embrace the open-source Deepseek model, claiming it poses an existential threat to OpenAI's business model.
In an interview with the South China Morning Post, Lee revealed his company has abandoned its previous strategy of training proprietary large language models, instead relying entirely on Deepseek's open-source offering. He stated that Deepseek's release triggered a "ChatGPT moment" in China, driving integration with the model by domestic hardware and software providers. This decision followed a surge in Chinese corporate demand for Deepseek in late January.
Lee believes Deepseek's free and open-source nature presents a fundamental challenge to OpenAI: "Sam Altman's worst nightmare is that his competitor is free. I've already met many people who canceled their ChatGPT subscriptions because Deepseek is free."
01.AI, with 200 employees, now plans to focus on customizing the Deepseek model for enterprise clients in finance, gaming, and legal sectors. Despite the strategic shift, Lee insists his company's technical expertise remains valuable, particularly in model training, fine-tuning, and inference. He projects Q1 2025 revenue to reach 100 million RMB ($13 million USD), equaling the company's entire 2024 revenue, though 01.AI is not yet profitable.
Regarding OpenAI and Anthropic's recent urging of the US government to ban the Deepseek model, Lee calls it a "display of paranoia": "They're watching their house of cards crumble because someone built an equally good house for free."
Lee also highlights the massive disparity in operating costs: OpenAI reportedly spent $7 billion in 2024, while Deepseek's cost is estimated at around 2% of that. He describes Deepseek as "infinitely sustainable," as its founders have sufficient funding and have reduced computational costs by 5 to 10 times.
