The Great AI Monopoly That Wasn’t
Predictions of AI-driven monopoly have outpaced the evidence. Several years into the generative AI boom, regulators have investigated, firms have invested, and markets have shifted—yet durable market power and demonstrable competitive harm remain elusive.
More than a year ago, Dirk Auer and I challenged the “hyperbolic and dystopian” narrative dominating discussions about competition in artificial-intelligence (AI) markets. We argued that concerns about competition risks—and calls for tougher enforcement—were unwarranted or overstated. AI markets were, in general, highly competitive. More importantly, the AI revolution offered an opportunity to revisit some of the assumptions driving competition concerns in digital services, including data as a barrier to entry and the role of network effects.
AI markets continue to evolve at breakneck speed. This makes it useful to revisit recent developments through the lens of economic and legal principles that inform competition policy. This post is the first in a biweekly series tracking those developments.
Rather than chase the news cycle, I will focus on concrete changes across the AI stack—compute, models, data, integration, and governance. The aim is to reassess, and where necessary push back on, common claims about competition, concentration, and market power. Over time, the series will examine infrastructure investment, talent mobility, model development, new products and services, partnership dynamics, and regulatory responses.
The goal is to connect real-world developments to the competition-policy and enforcement debate, and to test maximalist claims about inevitable concentration or competitive harm against observed market evidence.
The guiding question remains straightforward: Has the AI industry evolved toward monopolistic control of key inputs or demonstrable harm to competition, or toward layered, dynamic competition?