AI and the Logic of Arms Racing

This paper examines the emerging competition over artificial intelligence (AI) between the United States and China and evaluates the consequences of framing that competition as an "arms race." Policymakers in both countries increasingly describe AI as a domain of strategic rivalry tied to national power, military superiority, and regime security. This framing shapes budgetary priorities, military planning, and industrial policy, embedding AI within broader strategies of geopolitical competition. Yet AI differs from earlier strategic technologies. As a dual-use, general-purpose technology dependent on private-sector ecosystems, global supply chains, computing infrastructure, and data access, AI complicates traditional understandings of arms racing. To address this problem, the paper develops a typology of arms races and their likely political outcomes. The typology distinguishes among three forms of competition: latent arms races, which center on enabling resources such as semiconductors, computing power, industrial capacity, talent, and natural resources; conventional arms races, which focus on deployable military capabilities; and existential arms races, which frame competition as necessary for national or regime survival. The paper argues that these forms of competition produce distinct outcomes. Latent arms races are most closely associated with pathways to war, conventional arms races generate indeterminate outcomes ranging from escalation to stalemate, and existential arms races can paradoxically create incentives for cooperation and coordination. The analysis employs an analytically eclectic approach that combines realist and constructivist perspectives with historical case studies. Realist approaches explain the structural pressures driving competition over strategically significant technologies, while constructivist approaches illuminate how discourse and identity shape state behavior. Applying this framework to U.S. and Chinese AI strategies, the paper argues that contemporary AI competition fits all three categories but is manifested primarily as a latent arms race. The paper concludes that describing AI competition as an arms race is not analytically neutral; it is a performative act that helps produce the competitive dynamics it seeks to explain.