The methodology of selection in plant breeding has markedly evolved with the advent of high throughput molecular technology, the increasingly reasonable cost of genotyping, and the implementation of genomic selection (GS). For perennial crops, the potential of GS is high and gives the opportunity to shorten the breeding cycle by selecting at the juvenile stage using marker information. Here we present preliminary results of GS experiments for two perennials crop, Eucalyptus and oil palm, that play an important economical role in tropical regions. In the case of Eucalyptus, a simulation study was developed to test the efficiency of GS in the frame of a recurrent selection scheme for clone production over four breeding cycles. Scenarios crossing broad sense heritabilities (H²=0.6 and 0.1), dominance to additive variance ratios (R=0.1; 0.5 and 1) and training population structure were compared using Bayesian LASSO method. Models including dominance effects are all the more relevant when the R ratio and the training population size are high. The genetic gain per unit time with GS was 1.5 to 3 times higher than with phenotypic selection at mature stage for breeding and clone populations. For oil palm, we implemented a cross-validation approach with 111 individuals of the last generation of a key breeding population, evaluated through progeny tests including 40,000 individuals and genotyped with 140 microsatellites. The accuracy of GS increased when increasing the training population size and reached 0.6-0.7, according to the trait, with a 3:1 ratio for training and validation populations respectively. The small effective population size detected in this breeding population explains the good GS performance even with a limited panel of markers. Our studies based on two perennials crops presenting different biological patterns and different breeding contexts suggest very promising results of GS for long rotation plant species. (Texte integral)