Comment by raccoonone

Yep, we've been using GPUs for quite a while (even before the alpha support in Kube), both the K80s in Azure and some Pascals in our own clusters. With the support in Kube now it's quite seamless.