Becoming a State Project: Unpacking Power of Generative AI Companies

Generative AI presents a puzzle for political economy. Leading firms accumulate structural advantages, lock in users, and shape technical standards at unprecedented speed, while unit economics remain negative and no clear path to profitability has emerged. This puzzle, we argue, can only be made sense of through an explicit analysis of corporate power, for which mainstream frameworks centred on market concentration alone are ill-equipped. Drawing on heterodox economics and cultural political economy (Boyer, 2022; Galbraith, 1984; Rothschild, 2002; Sum and Jessop, 2013), we develop a multi-dimensional heuristic distinguishing market power, as strategic control within markets, from the power to shape the wider cultural political economy. The rules of the game and the relationship to the state. We map market concentration across three layers of the genAI stack (GPU infrastructure, hyperscalers, foundation models), examine its distinctive cost structure, and analyse the emerging state–capital configuration. High costs and negative unit economics generate strong concentration, pointing toward an AI oligopoly. Politically and culturally, firms deploy the familiar Big Tech playbook (lobbying, academic capture, hegemony production), recruited around two narratives: AI nationalism and the AGI imaginary. Yet we identify a structural break from the platform era. Where platform firms sought to bypass the state, frontier AI firms actively court state procurement and patronage. What is emerging, we argue, is a state project in the making: a configuration in which states adopt the survival and dominance of specific AI firms as their own objectives.