This paper studies inference performance of Instrumental Variables Estimators in situations where error terms are heteroskedastic and there are many instruments. In particular, performance of a estimator proposed by Hausman, Newey, Woutersen, Chao, and Swanson (2007) with the robust version of JIVE -proposed by Angrist, Imbens and Krueger (1999)- is analyzed. Theoretical results are presented for the robust t-statistics, which is mostly affected by the finite-sample bias of the estimator.