risk · R06

AI Power Concentration and Oligopoly Control

The top AI companies are consolidating power through unprecedented funding rounds (e.g., OpenAI's $110B), pushing frontier model development costs towards $1B+. This creates insurmountable barriers to entry and concentrates technological and policy-making power in a handful of unelected entities.

Updated 2026-09-20

Observatory assessment: 88 / 100. High

How assessments are made

Deep dive

AI development is consolidating into a narrow oligopoly at a pace that exceeds even the tech monopolies of the social media era. Training a frontier model now costs $100M-$1B, creating insurmountable barriers to entry. The compute required doubles roughly every 6-10 months, meaning only companies with access to massive GPU clusters (primarily NVIDIA H100/B200) can compete.

This concentration has downstream effects: these companies control the API layer through which most businesses access AI, creating platform dependency. They also control the training data pipelines, model architectures, and increasingly the hardware supply chain (custom chips). The 'kingmaker' dynamic means a single company's policy decisions—about content filtering, pricing, or capability restrictions—effectively become global AI policy without democratic input.

Trajectory and projections

Mitigation strategies

• Public investment in open-source foundation models (EU/CERN-style)
• Antitrust enforcement targeting vertical integration in AI stack
• Mandatory API interoperability standards for AI platforms
• Compute access programs for researchers and developing nations
• Proposals for nationalization of key AI labs to ensure public oversight

Top evidence