With the advent of next generation sequencing technologies, alignment and phylogeny estimation of datasets with thousands of sequences is being attempted. To address these challenges, new algorithmic approaches have been developed that have been able to provide substantial improvements over standard methods. This paper focuses on new approaches for ultra-large tree estimation, including methods for co-estimation of alignments and trees, estimating trees without needing a full sequence alignment, and phylogenetic placement. While the main focus is on methods with empirical performance advantages, we also discuss the theoretical guarantees of methods under Markov models of evolution.Finally, we include a discussion of the future of large-scale phylogenetic analysis.