In decadal predictability studies, the subpolar Atlantic stands out as a region of high potential and real predictability. Since local temperature and salinity variations in the region are for a large part controlled by ocean dynamics, skillful predictability of the local ocean dynamics is a prerequisite to obtain multiyear predictability of other variables such as sea surface temperature. In this study, we discuss the predictability of the main ocean current system in the region, the subpolar gyre. From perfect model hindcasts exploiting initial condition information only from realistic ocean observation locations, we find that predictability is increased when Argo subsurface data are included. In our real-world experiments with initialized hindcasts, the observed decline in subpolar gyre strength of the mid-1990s is reproduced well and we find predictability of the subpolar gyre up to 2 years ahead, comparable to the skill of a damped persistence model.