Baidu’s CEO Robin Li has expressed skepticism about the sustainability of open-source AI models, particularly highlighting DeepSeek’s recent release. Li argues that while open-source models may currently match proprietary models in capability, maintaining competitiveness will become increasingly expensive as AI development progresses. He estimates that by 2025, the cost to develop and train competitive large language models could reach $100 million, making it financially challenging for open-source projects to keep pace. Li’s comments reflect growing industry debate about the viability of open-source AI versus proprietary models, especially considering the massive computational resources and financial investments required. The Baidu chief emphasizes that their own AI model, Ernie, will remain proprietary to maintain competitive advantage. His statements come amid increasing competition in the AI sector, where companies must balance innovation costs with market strategy. The article also notes that while open-source models currently show promise, the escalating costs of AI development may force a shift toward more proprietary approaches. This perspective challenges the open-source movement in AI, suggesting that economic realities could limit its long-term impact on the industry. The discussion highlights the complex interplay between technological advancement, financial sustainability, and market dynamics in the evolving AI landscape.
Source: https://www.businessinsider.com/baidu-ceo-deepseek-open-source-model-ernie-cost-earnings-2025-2