Combining multiple displacement amplification (MDA) with metagenomics enables analysis of environmental or host-associated samples with extremely low DNA concentrations that would otherwise not be appropriate for high-throughput sequencing. Although amplification bias and non-specific amplification have been reported from MDA-amplified samples, the impact of MDA on metagenomic datasets is not well understood. In this study, we compared three MDA methods (i.e., bulk MDA, emulsion MDA, and primase MDA) for metagenomic analysis of two different picogram-range DNA template concentrations (approx. 1 and 100 pg) derived from a microbial community standard “mock community” and two low biomass environmental samples (i.e., borehole fluid and groundwater). We assessed the impact of MDA on metagenome-based community composition, assembly quality, functional profiles, and binning. We found amplification bias against high GC content genomes but relatively low non-specific amplification such as chimeras, artifacts, or contamination for all MDA methods. We observed MDA-associated representational bias for microbial community profiles, especially for low-input DNA and with the primase MDA method. Nevertheless, similar taxa were represented in MDA-amplified libraries to those of unamplified samples. The MDA libraries were highly fragmented, but similar functional profiles to the unamplified libraries were obtained for bulk MDA and emulsion MDA at higher DNA input and across these MDA libraries for the groundwater sample. Medium to low-quality bins were possible for the high input bulk MDA metagenomes for the most simple microbial communities, borehole fluid, and mock community. Overall, our results suggest that MDA-based amplification should be avoided, but that it is nonetheless possible to access meaningful taxonomic and functional information from samples with extremely low DNA concentration where direct metagenomics analysis would otherwise not be feasible.