Near infrared spectroscopy (NIRS) has been widely used in analyzing various real complex samples due to its superiority in fast and non-destructive determination. It is difficult, however, for the technique to analyze the components of micro-content, especially the inorganic components, because NIR spectrum contains only the weak signals of C—H, O—H and N—H, and inorganic components generally generate no response in the spectrum. In this review, the principle and applications of NIRS in analyzing inorganic components in environmental, soil, plant and bio-samples are summarized. The feasibility of applying NIRS in micro-inorganic analysis was proved by various applications of the technique in analyzing lake or river sediments, forest or farmland soils, straw, amaranth, clover, paprika, rice, wheat, meats, and sea foods. Because multivariate calibration technique is generally adopted in NIR spectral analysis, quantitative analysis can be achieved by modeling the spectral responses in which the interactions of the inorganic components and organic contents are included or by using the correlationship between the micro and macro components. However, the models must be used with care in such applications because they may fail when the predicted samples are heavily deviated from the calibration ones. Furthermore, applications of preconcentration techniques in NIR analysis are also summarized in this paper. By adsorption of an inorganic analyte onto the surface of a high efficient adsorbent from a dilute solution, the interactions between the analyte and the adsorbent can be directly measured by using NIR diffuse reflectance spectroscopy (NIRDRS), and the relationship between the contents of the analyte and the spectra can be modeled with multivariate calibration techniques. The adsorbents of functionalized resins, thiol-functionalized magnesium phyllosilicate clay, nano-hydroxyapatite, etc. have been studied for detection of micro inorganic ions in dilute solutions such as Hg2+, Ag+, Pb2+, Zn2+, Cu2+, Co2+, Ni2+, Cd2+ and Cr3+. The results of these studies showed that the adsorption can significantly improve the detection limit and fast detection can be achieved by NIRDRS measurements and multivariate calibration.