We propose a high dimensional minimum variance portfolio estimator under statistical factor models, and show that our estimated portfolio enjoys sharp risk consistency. Our approach relies on properly integrating l1 constraint on portfolio weights with an appropriate covariance matrix estimator....