Comments on deconfinement in AdS/CFT

We study the index of $\mathcal{N}=4$ Yang-Mills theory on $S^3\times\mathbb{R}$. We argue that the index should undergo a large $N$ deconfinement phase transition, by computing an upper bound of its `temperature.' We compute this bound by optimizing the phases of fugacities. The bound we find has some features analogous to the Hagedorn temperature. We briefly discuss a possible mechanism of the actual deconfinement transition below our bound. Our upper bound is lower than the Hawking-Page transition `temperature' of known BPS black holes in the AdS$_5$ dual. We thus expect the existence of new black holes.