Concentrating intelligence: scaling and market structure in artificial intelligence.

This paper analyzes the competitive dynamics and market structure of AI foundation models, with a focus on large language models (LLMs) such as OpenAI's GPT-4 and Google DeepMind's Gemini. It highlights the significant economies of scale and scope driven by high fixed costs in pre-training, reliance on key inputs like computational resources, proprietary data, and specialized talent, which contribute to market concentration risks and potential natural monopolies. The study discusses concerns about market tipping, vertical integration by dominant technology firms, and the implications for competition policy, including data sharing, interoperability, and merger scrutiny. It also addresses the challenges of balancing competition with AI safety, regulatory frameworks—particularly in the European Union—and the broader societal stakes of AI governance as these models become increasingly integral to the economy.