OpenAI had pioneered the AI revolution, and is still the most valuable AI startup, but some experts believe that it might find it hard to maintain its advantage.
Kai-Fu Lee, a prominent AI pioneer and computer scientist, recently shared some compelling insights into the evolving landscape of large language models (LLMs). His comments highlight the increasing dominance of open-source models and the potential challenges faced by companies like OpenAI that have invested heavily in closed-source development. He suggests a potential existential threat to OpenAI’s business model, hinting at sleepless nights for OpenAI CEO Sam Altman.
“In both the US and China, the pre-training of a giant model has consolidated and is consolidating,” Lee said. “And it’s become clear that open source will be the winner,” he claimed.
He acknowledged that some companies are resistant to this idea. “There’s still many that will not concede, right? OpenAI and Anthropic who build their businesses believing they can build a better closed model than everyone else.”
Lee believes these companies were taken aback by the rapid advancements in open-source alternatives. “I think they got shocked when they saw a model as good,” he observed. He then pointed to the significant cost disparity between open-source and closed-source development: “If you think about OpenAI’s costs of $7 billion of operating costs in 2024, DeepSeek probably operated with 2% the operating expense.”
This cost difference, according to Lee, is the crux of the issue. “So the issue really isn’t whose model is one percent better,” he explained. “I think they’re all very good. But the issue is, is OpenAI’s model even sustainable, right?” He emphasizes the financial strain on OpenAI: “I mean you [are] spending seven billion dollars a year, making a massive loss.” Meanwhile, “here you have a competitor coming in with an open-source model that’s for free.” Furthermore, he emphasizes the long-term viability of open-source competitors like DeepSeek: “and that company also is infinitely, infinitely lasting because this founder has enough money to fund it at the current level and has reduced the cost of computing by a factor of 5 to 10.”
Lee’s stark assessment concludes with a pointed jab at OpenAI’s leadership: “So with that kind of formidable competitor, I think Sam Altman is probably not sleeping well.”
Lee’s comments come at a time of change in the AI industry. The rapid rise of open-source models like DeepSeek and Meta’s Llama have cut away at some of the advantages of closed labs. These models often offer comparable performance to their closed-source counterparts at a fraction of the cost, empowering researchers and developers with unprecedented access to powerful AI tools. This accessibility fosters innovation and collaboration, potentially accelerating the overall progress of the field. OpenAI’s substantial operating costs, coupled with the increasing competitiveness of free and open-source models, raise serious questions about the long-term viability of its current business model. While OpenAI still holds a significant brand advantage and strong partnerships — and its models are leading most benchmarks — the pressure from open-source competitors is undeniably mounting, potentially forcing a strategic shift in the near future. And with AI veterans like Kai-Fu Lee saying that open-source models are giving OpenAI sleepless nights, OpenAI might need to evaluate the open-source threat before it ends up disrupting its business model.