Marine sponges harbour complex microbial communities with significant biosynthetic potential, yet their diversity may be underestimated by conventional short-read sequencing approaches. We applied Nanopore long-read metagenomic sequencing to characterise the microbiomes of Geodia barretti (High Microbial Abundance, HMA) and Mycale lingua (Low Microbial Abundance, LMA), two ecologically distinct sponge species from Norwegian reef ecosystems. Metagenomic assembly using MetaFlye generated high-quality assemblies totalling 253 Mb (G. barretti) and 235 Mb (M. lingua) with N50 values of 30.6 kb and 7.7 kb respectively. Taxonomic profiling using DIAMOND-MEGAN protein-based alignment revealed substantially greater microbial diversity compared to k-mer based methods, detecting bacterial sequences comprising 81.4% of the G. barretti metagenome, dominated by Proteobacteria, Chloroflexi, and Candidatus Poribacteria. The HMA sponge exhibited remarkable taxonomic complexity spanning 13 bacterial phyla, whilst the LMA M. lingua showed reduced bacterial abundance (15.8%) as expected. Biosynthetic gene cluster analysis using antiSMASH identified 153 BGCs in G. barretti, representing exceptional biosynthetic richness including 62 terpene, 39 RiPP, and 14 NRPS clusters. BiG-SCAPE network analysis revealed diverse gene cluster families with substantial singleton clusters, indicating previously uncharacterised biosynthetic potential. In contrast, M. lingua harboured minimal BGC content, consistent with its low microbial load. Our findings demonstrate that long-read sequencing combined with protein-based taxonomic profiling enables more comprehensive characterisation of complex host-associated microbiomes. This approach reveals marine sponges as rich reservoirs of microbial diversity and cryptic biosynthetic pathways, providing a valuable framework for future bioprospecting and natural product discovery from marine holobionts.