Market Power

Definition

Market power in the AI context refers to the capacity of a small number of dominant firms to exert disproportionate control over the conditions under which AI technologies are developed, distributed, and accessed. This power derives from control over scarce resources including massive training datasets, specialized computing infrastructure, top research talent, and proprietary model architectures. Market power enables dominant firms to set prices, determine API access terms, establish technical standards, and influence the strategic direction of AI development. Unlike traditional market concentration, AI market power is reinforced by data network effects — the more data a firm controls, the better its models perform, attracting more users and generating more data in a self-reinforcing cycle.

How It Relates to AI Threats

Market power is a core concern within the Economic and Labor Threats domain. Under the power and data concentration sub-category, the concentration of AI capabilities among a small number of firms creates multiple threat vectors. Dependent organizations face pricing vulnerability, lock-in, and the risk of service discontinuation without viable alternatives. Startups and smaller firms face barriers to entry that limit competitive innovation. The concentration of AI infrastructure creates systemic risk, as disruptions to a dominant provider propagate across all dependent services. Market power also enables dominant firms to shape regulatory outcomes through lobbying, to set de facto standards that entrench their position, and to influence the trajectory of AI research toward commercially advantageous directions.

Why It Occurs

• Training frontier AI models requires compute infrastructure investments exceeding the capital available to all but the largest firms
• Data network effects create self-reinforcing advantages where more users generate more data, improving models and attracting still more users
• Talent concentration at a small number of firms limits the diffusion of cutting-edge AI research capability across the broader economy
• Cloud infrastructure dependencies mean that AI services from competing providers often run on platforms controlled by dominant firms
• Intellectual property protections and trade secrets around model architectures and training methodologies limit knowledge transfer

Real-World Context

While no specific incidents in the TopAIThreats taxonomy currently document market power abuses, the structural conditions are well-documented. A small number of firms — including those developing the most capable foundation models — control significant shares of AI compute, training data, and research talent. Competition authorities in the EU, US, and UK have opened investigations into AI market concentration, cloud computing market power, and partnership structures that may circumvent merger review. The EU AI Act, Digital Markets Act, and proposed US legislation address aspects of AI market power, while international bodies including the OECD have published analyses of competition dynamics in AI markets.