Most microbial detection techniques require pretreatment, such as fluorescent labeling and cultivation processes. Here, we propose novel tools for classifying and identifying microorganisms such as molds, yeasts, and bacteria based on their intrinsic dielectric constants in the THz frequency range. We first measured the dielectric constant of films that consisted of a wide range of microbial species, and extracted the values for the individual microbes using the effective medium theory. The dielectric constant of the molds was 1.24-1.85, which was lower than that of bacteria ranging from 2.75-4.11. The yeasts exhibited particularly high dielectric constants reaching 5.63-5.97, which were even higher than that of water. These values were consistent with the results of low-density measurements in an aqueous environment using microfluidic metamaterials. In particular, a blue shift in the metamaterial resonance occurred for molds and bacteria, whereas the molds have higher contrast relative to bacteria in the aqueous environment. By contrast, the deposition of the yeasts induced a red shift because their dielectric constant was higher than that of water. Finally, we measured the dielectric constants of peptidoglycan and polysaccharides such as chitin, α-glucan, and β-glucans (with short and long branches), and confirmed that cell wall composition was the main cause of the observed differences in dielectric constants for different types of microorganisms.