Abstract: Non-dispersive infrared (NDIR) CO₂ gas sensors operate on the principle of selective infrared absorption by CO₂ molecules at specific wavelengths. These sensors offer high selectivity, low operating temperatures, a wide measurement range, excellent stability, a compact structure, low maintenance costs, and a long service life. They enable online detection of CO₂ concentrations in complex environments and show broad application prospects in agricultural production, healthcare, and environmental monitoring. This work reviews the current status of infrared sources, detectors, and optical gas chambers for NDIR CO₂ sensors, providing an in-depth analysis of technological advancements in fully integrated miniature systems. Furthermore, the study systematically examines self-compensation strategies from both software and hardware perspectives, while discussing associated intelligent machine learning algorithms. Finally, future development trends are outlined, highlighting the progression of CO₂ infrared gas sensors toward high precision, wafer-scale integration, and intelligent capabilities.