Unsupervised classification of digital pathology images in glomerulonephritis has the potential to reveal novel pathological features associated with disease severity. In this study, we aimed to develop an artificial intelligence (AI)-based deep learning model for clustering pathology image patches and to correlate these clusters with various clinical features of glomerulonephritis. We analyzed over 10,000 pathology image patches derived from kidney biopsies of 40 patients with glomerulonephritis. Utilizing the Conchi foundation model—which integrates stain normalization with a transformer-based approach—we extracted features from the image patches and subsequently performed K-means clustering. The relative proportions of the resulting clusters were then correlated with kidney function parameters for each patient. Additionally, AI-driven clustering outcomes were compared with annotations provided by pathologists. Our unsupervised analysis yielded 100 distinct clusters, effectively distinguishing major kidney microstructures such as glomeruli and vascular structures. Notably, the clustering approach also identified specific pathological features, including tubulointerstitial fibrosis. Several clusters were significantly associated with clinical markers of kidney dysfunction, such as high levels of proteinuria and reduced estimated glomerular filtration rate. Furthermore, the clusters derived from the AI model exhibited strong concordance with pathologist-driven annotations. The results demonstrate that AI-based unsupervised clustering of kidney biopsy images can effectively differentiate major renal microstructures and pathological features. This approach offers a rapid and efficient method for the analysis of digital kidney pathology images. Ongoing studies aim to extend these findings by integrating unsupervised disease prediction clustering with spatial transcriptomics to further elucidate the clinicopathological significance of the identified clusters