Dynamic AI Compute Demand Model (DACDM): Complexity Migration, Cognitive Commoditization, and the Dynamics of AI Infrastructure Demand

Dynamic AI Compute Demand Model (DACDM) is a conceptual framework for analyzing AI compute demand as a dynamic interaction between workload growth, task complexity, model and hardware efficiency, complexity migration, and substitution across execution pathways. Rather than assuming that growth in AI usage translates proportionally into growth in accelerator demand, DACDM examines how previously difficult tasks may migrate toward cheaper and more efficient execution pathways as models, hardware, routing, caching, distillation, and other forms of compression improve. This v1.0 release presents the conceptual framework and its testable hypotheses. It is not an empirically validated forecasting model, and no claim of empirical support is made at this stage. The accompanying DACDM Pilot 01 — Code Generation Complexity Compression protocol defines a frozen, pre-registered empirical test of H2–H4 in the code-generation domain. The pilot is designed to test whether complexity migration and compression beyond hardware efficiency can be observed and falsified under controlled conditions. Author: CHAU HUNG SAN (辛秋雄)Version: 1.0DOI: 10.5281/zenodo.21930795