Scaling AI Performance & Chiplet Architectures