Dragon fruit has been identified as a key strategic crop in Binh Thuan, Vietnam. However, dragon fruit farmers in this area are exposed to many risks impacting their livelihoods and agricultural productivity. This study, therefore, investigates the sources of risk these farmers are facing and explores heterogeneity in risk perception by segmenting farmers into distinct clusters. This study employed clustering methods and specifically mixes a hierarchical approach and a Partitioning Around Medoids (PAM) approach, to cluster the 315 dragon fruit farmers. Next, a multinomial logistic regression model was used to examine how clusters were associated with socio-demographic and farm characteristics. Risks were classified in eight domains: weather, biological, supply, demand, financial, logistic, operational and policy risks. While demand, biological, and supply risks are perceived to be high, logistic and policy risks are perceived low. More specifically, demand instability, output price volatility and pest and disease outbreaks were perceived as most important. Three clusters were found: Cluster 1 with a "high risk perception", Cluster 2 forms a "medium risk perception" group, and Cluster 3 with "low risk perception". The multinomial regression results showed that educational level and residence were significantly associated with both the high and medium risk perception clusters. In comparison to the medium risk perception group, the high risk perception group was negatively and significantly associated with larger household size, greater farming experience, and adherence to certification standards. The diverse risk profiles of the farmers could be used to support the effective design of tailored risk management policies.