Background. Complex carcinogenic mechanisms and the existence of tumour heterogeneity in multiple myeloma (MM) prevent the most commonly used staging system from effectively interpreting the prognosis of patients. Since the microenvironment plays an important role in driving tumour development and MM occurs most often in middle-aged and elderly patients, we hypothesize that ageing of bone marrow mesenchymal stem cells (BM-MSCs) may be associated with the progression of MM. Methods. In this study, we collected the transcriptome data on MM from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO) databases. Differentially expressed genes in both senescent MSCs and MM tumour cells were considered relevant damaged genes. GO and KEGG analyses were applied for functional evaluation. A PPI network was constructed to identify hub genes. Subsequently, we studied the damaged genes that affected the prognosis of MM. Least absolute shrinkage and selection operator (lasso) regression was used to identify the most important features, and a risk model was created. The reliability of the risk model was evaluated with the other 3 GEO validation cohorts. In addition, ROC analysis was used to evaluate the novel risk model. An analysis of immune checkpoint-related genes, tumour immune dysfunction and exclusion (TIDE), and immunophenotypic scoring (IPS) were performed to assess the immune status of risk groups. pRRophetic was utilized to predict the sensitivity to administration of chemotherapeutic agents. Results. We identified that MAPK, PI3K, and p53 signalling pathways were activated in both senescent MSCs and tumour cells, and we also located hub genes. In addition, we constructed a 14-gene prognostic risk model, which was analysed with the ROC and validated in different datasets. Further analysis revealed significant differences in predicted risk values across the International Staging System (ISS) stage, sex, and 1q21 copy number. A high-risk group with higher immunogenicity was predicted to have low proteasome inhibitor sensitivity and respond poorly to immunotherapy. Lipid metabolism pathways were found to be significantly different between high-risk and low-risk groups. A nomogram was created by combining clinical data, and the optimization model was further improved. Finally, real-time qPCR was used to validate two bortezomib-resistant myeloma cell lines, and the test confirmed that 10 genes were detected to be expressed in resistant cell lines with the same trend as in the high-risk cohort compared to nonresistant cells. Conclusion. Fourteen genes related to ageing in BM-MSCs were associated with the prognosis of MM, and by combining this genotypic information with clinical factors, a promising clinical prognostic model was established.