In order to show an application of GARCH models to exchange rates, we used statistical techniques such as principal component analysis and multivariate time series analysis to model mean and variance (volatility). The use of principal component analysis helps to reduce the dataset size and lead to fit fewer models, without losing original set information. The use of GARCH models is justified by the presence of heteroskedasticity on the exchange rates returns variance. Based on the fitted models new daily series were simulated, using Monte Carlo method (MC), and used to create confidence interval estimates for exchange rates future scenarios. For the proposed application were chosen exchange rates with bigger market share according to the BIS study, released every three years.