Forest or land fires are events that often occur in various countries in the world that require serious handling from all parties because they have an impact on all lines of life. Therefore, early treatment is needed, one of which is by grouping fire-prone areas using hotspot data. Hotspots can be obtained by satellite in this study taking hotspots from NASA satellites. The data used are latitude, longitude, brightness and confidence. The method used is the Clara Clustering method because this method has the advantage of being resistant to outliers and can be used in large amounts of data. This study concludes that the Clara method can process 16,579 hotspot data with the best Shilhoutte Coefficient value of 0.89 with 2 clusters while the potential possessed by cluster 1 is included in high potential with an average brightness of 3670K and a confidence value of 1 or hight. Meanwhile, cluster 2 has moderate potential with an average brightness of 3490K and a confidence value of 3 or nominal.