Artificial Intelligence as Cognitive Infrastructure: Beyond the Tool Paradigm

The dominant classification of Artificial Intelligence as a tool is both historically insufficient and analytically imprecise. This paper proposes a more consequential reclassification: AI as cognitive infrastructure. Drawing on five converging analytical lenses -- historical analogy, dependency analysis, cognitive offloading theory, invisible ubiquity, and institutional governance response -- the paper examines whether AI has crossed the threshold from tool to infrastructure. Unlike tools, which are optional, external, and bounded, infrastructure is load-bearing, ambient, and structural. Through analysis of AI's role in healthcare, education, finance, and governance -- supported by empirical deployment data -- this paper argues that AI increasingly exhibits the defining characteristics of cognitive infrastructure as identified by Star and Ruhleder. Grounded in Clark and Chalmers' Extended Mind thesis, the paper further contends that AI is no longer merely a system that assists human reasoning, but increasingly one through which reasoning is conducted. The implications are significant for policy, equity of access, governance, and AI literacy.