Background: Thiazolidinedione (TZD) exerts anti-proliferative effects on multiple myeloma (MM) cells. However, there has not been any human study investigating the risk of MM associated with TZD use. Methods: We used Taiwan’s National Health Insurance database to identify 423,949 patients who had been newly diagnosed with diabetes mellitus between 1999 and 2005. After excluding ineligible patients, 86,999 pairs of patients with and without the use of TZD (rosiglitazone or pioglitazone) that had been matched based on propensity score were selected for a follow-up for MM until 31 December 2011. The hazard ratios for MM were estimated using Cox regression and weighted using a propensity score. Results: After a median follow-up of 4.6 years and 4.7 years in ever users and never users of TZD, 32 and 47 cases were diagnosed with MM, respectively. A 35% lower risk (though not statistically significant) was observed among ever users (hazard ratio 0.652, 95% confidence interval: 0.416–1.023, p = 0.0625). When ever users were divided by the median (15 months) cumulative duration of TZD therapy, the hazard ratios (95% confidence interval) for the lower and upper medians were 0.706 (0.394–1.264) and 0.603 (0.346–1.051), respectively. When treated as a continuous variable, the hazard ratio for every 1-month increment of the cumulative duration was 0.980 (95% confidence interval: 0.963–0.997, p = 0.0185). In the age subgroup analysis, a significantly lower risk could be seen in the older age subgroup of ≥65 years (hazard ratio 0.550, 95% confidence interval: 0.305–0.992, p = 0.0468). Additional analyses suggested that there were no interactions between TZD and some medications and between TZD and some clinical diagnoses, and that the use of TZD as a preventive drug for MM might not be cost-effective because a number-needed-to-treat of 5800 was too large. Survival analyses suggested that ever users had a significantly lower risk of death when all patients were analyzed (hazard ratio: 0.84, 95% confidence interval: 0.81–0.87, p < 0.0001 via a log-rank test) or when patients who developed MM were analyzed (hazard ratio: 0.40, 95% confidence interval: 0.19–0.86, p = 0.0153 via a log-rank test). Conclusions: In Taiwanese patients with type 2 diabetes mellitus, TZD use is associated with a borderline lower risk of MM, which is more remarkable in patients aged ≥65 years. Because of the low incidence of MM, the use of TZD for the prevention of MM may not be cost-effective. Patients who have been treated with TZD may have a survival advantage. Future research is required to confirm the findings.