Semiconductor manufacturing requires production decisions across multiple hierarchical levels under complex, stochastic, and highly dynamic operating conditions. Simulation (SIM), optimization (OPT), and artificial intelligence (AI) have been widely applied to support these decisions, while recent studies increasingly combine these approaches to exploit their complementary capabilities. This study systematically reviews 92 publications from 2021 to approximately June 2026 on the application and integration of SIM, OPT, and AI for production decision-making in semiconductor manufacturing. The reviewed studies are classified into four production decision problem categories—planning, capacity allocation, scheduling and dispatching, and production performance prediction and optimization—and seven methodological configurations: SIM, OPT, AI, SIM+OPT, SIM+AI, OPT+AI, and SIM+OPT+AI. Quantitative analysis shows that scheduling and dispatching is the dominant research area, AI is the most frequently used stand-alone approach, and hybrid approaches account for a substantial share of recent studies. Beyond frequency-based analysis, this review examines the functional roles, information flows, and recurring integration mechanisms of the three methodologies. The findings show that SIM primarily supports system representation and evaluation, OPT provides prescriptive decision generation, and AI enables prediction and adaptation. Based on these findings, five research directions are identified, leading toward closed-loop production decision intelligence that integrates representation, prediction, optimization, execution, evaluation, and learning in semiconductor manufacturing.