Soil microbes exert critical control over ecosystem functions by regulating organic matter decomposition and nutrient cycling. Diverse methods have been adopted to estimate the soil microbial biomass carbon (C_mic) and nitrogen (N_mic), largely based on fumigation and respiration or biomarkers, such as phospholipids and ATP. Although soil microbial DNA is also an important biomarker, whether this biomarker can be used to estimate C_mic and N_mic at a large scale in arid and semi-arid regions that remains unclear. To explore this, the 124 soil samples collected from 28 different vegetation/microhabitat types in diverse ecosystems (i.e., desert, cropland, typical steppe, meadow steppe and forest) across arid and semi-arid regions in northern China; then, the strong linear relationships were developed among soil microbial DNA concentrations with the C_mic and N_mic contents estimated by the classic chloroform fumigation-extraction (CFE) method. The results showed that the soil microbial DNA concentrations were linearly related to the observed C_mic (r = 0.99) and N_mic (r = 0.97) across the diverse vegetation types, although soil microbial biomass presented an enormous fluctuation. Moreover, based on the above-mentioned coefficients of the linear relationship, it was further inferred that the mean proportions of carbon and nitrogen elements in DNA were 0.99% and 3.70% for C_mic and N_mic, respectively, when assuming the G-C base-pair proportion is 50% in double-stranded DNA. In conclusion, the study proved that soil microbial DNA as a powerful indicator may provide convenient and efficient estimates for C_mic and N_mic in arid and semi-arid systems. Although this method still needs to be further improved and examined at a larger scale, it provides a good addition to the already existing methods to determine microbial biomass.