Two Structures of AI Power: National Capacity, Relational Boundary Control, and Systemic Influence in the Emerging AI Order

<jats:p>Artificial intelligence competition is often measured through models, compute, investment, and adoption, yet these indicators conflate resources, relational leverage, and system-level outcomes. This working paper develops a two-structure framework that distinguishes AI National Power from AI Boundary Power. AI National Power is the usable capacity of a national system to generate, mobilize, govern, and sustain AI-relevant resources. AI Boundary Power is an actor-level relational capacity to alter the conditions under which another actor can obtain, transfer, integrate, deploy, or continue using an AI-relevant resource at a specified interface. The paper further separates boundary control from target-side boundary autonomy and from systemic influence as a later outcome. Through mechanism illustrations involving advanced-computing export controls and European Union AI regulation, it argues that national depth and boundary control are distinct but potentially complementary forms of AI power.</jats:p>