Improving AI inference at the edge For peak performance at the edge, it’s critical to scale storage down to a single node, in order to bring inference closer to the data.