Artificial Intelligence, Competition, and Welfare

We propose a policy-relevant research agenda examining how market power in upstream artificial intelligence (AI) affects downstream prices, industry structure, factor returns, and welfare—especially whether labor-displacing AI leaves workers worse off. In our open-economy general equilibrium model, AI is a priced, imported input. Distributional effects depend on how sectoral skill intensity responds to AI prices. Non-monotonicity can cause “double harm” for displaced workers, who may face lower real wages both when AI is cheap and again if prices rise due to market power. Our main model features two non-traded sectors and firms making discrete adoption decisions about technology. Adoption reduces unit costs, displaces some types of workers, and depresses wages for those workers via diminishing returns elsewhere, while leaking AI fees abroad. Strategic AI pricing reduces welfare by raising downstream marginal costs (via usage fees) and limiting entry and variety (via access fees). We derive an adoption frontier linking feasible usage fees to displaced workers’ outside options, showing that a monopolist typically prices on this boundary; capping one fee shifts rents to the other. Regulating both fees, alongside policies that absorb displaced labor, can raise national welfare.