Competition at Risk in Generative AI Markets

The good news is that global competition between AI models has intensified significantly since 2024. A "winner-takes-most" market has not yet emerged. Different model providers compete with each other, including through open-source approaches.

"Competition at the model level is currently working," explains cep digital expert Anselm Küsters, who co-authored the study with cep economist Matthias Kullas. "But at other levels of the value chain, we continue to observe concentration and tendencies to restrict competition." Above all, access to sufficient computing power remains critical for market access and is increasingly exacerbated by an energy bottleneck. Those who cannot secure access to electricity, grids and sufficient energy today will fall behind in the race to build the next generation of frontier models. Risks are also increasing in the downstream markets for AI services. The integration of generative AI into platforms and the use of AI agent systems increase dependencies and switching costs.

A core problem lies at the institutional level. "Antitrust proceedings are too slow for exponential technologies," warns Matthias Kullas. "If decisions are only made after years, they come too late."

The cep calls for the existing instruments to be applied more quickly and consistently to the AI market in the course of 2026 in order to prevent a permanent consolidation of market structures and dependencies.