Hugging Face's Clem Delangue on Open Source AI and the LLM Bubble | MTS Live

What It Covers

Hugging Face CEO Clem Delangue argues that open source AI accelerates safety rather than undermining it, while warning of an LLM API bubble and positioning robotics as AI's next major frontier.

Key Questions Answered

• •Open Source vs. Closed APIs: Chinese organizations including DeepSeek, Qwen, and Qimi now dominate open source AI contributions while US frontier labs retreat behind closed APIs. Most US startups and academic researchers currently rely on Chinese open source models, reversing America's historical leadership position.
• •LLM Bubble Risk: Overinvestment concentrates specifically in large language models distributed behind closed APIs, not AI broadly. Massive data center buildouts continue despite uncertain long-term margins and unclear competitive moats, making this segment the most vulnerable to correction in the near term.
• •Open Source as a Security Asset: Restricting model access creates asymmetric risk where attackers gain capabilities defenders lack. Broad open access lets defenders build protection systems in parallel with offensive tools, making the overall ecosystem more resilient than a closed model controlled by few players.
• •Robotics Scale Signal: Hugging Face's LeRobot shipped nearly 10,000 units globally in roughly one year, with over 300 third-party apps already built for the platform. Delangue identifies Chinese manufacturers as current robotics leaders, urging US startups to accelerate hardware development using existing frontier model strengths.